技术领域Technical Field
本发明涉及流场处理技术领域，特别涉及基于物理信息融合的流场特征提取跟踪方法、装置、设备。The present invention relates to the technical field of flow field processing, and in particular to a flow field feature extraction and tracking method, device and equipment based on physical information fusion.
背景技术Background Art
针对三维时变流场，由于其数据量大、数据结构复杂且与视点相关，在匹配效率、交互可视化等方面仍存在较多问题亟需解决。通过跟踪流场中这些变化的特征，如涡、激波等，研究者可以更深入地理解流体动力学现象。这对于揭示湍流、涡旋、边界层等流场结构的演变和相互作用有重要意义。特征跟踪通过在相邻时间步长之间匹配特征元素，分析流场特征的演化历程，从而帮助研究人员定义特征事件的发生，例如，延续、分裂、合并等，事件检测就是为了描述特征所发生的发展和变化。因此，为了探索和分析流场的演化过程，特征跟踪及其结果可视化是流场科学可视化的重要任务。现有经典的属性向量算法和体积重叠算法能解决一定程度上的特征匹配问题，然而时变流场数据庞大，经典算法在运用相似性度量对特征进行匹配时需要依赖用户对各属性设置的阈值进行判断，要求用户具有丰富的领域先验知识，其中，等值面方法是用于可视化阈值区域表面的最常用的技术。但计算机视觉领域的方法大多针对二维图像数据，对于三维时变流场而言，由于其数据量大、数据结构复杂且与视点相关，在匹配效率、交互可视化等方面仍存在较多问题亟需解决。为此，需要设计更加便于实现的特征跟踪方法帮助有需求的科研人员进行特征在时空域上的行为分析。For three-dimensional time-varying flow fields, due to the large amount of data, complex data structure and viewpoint-related, there are still many problems in matching efficiency and interactive visualization that need to be solved. By tracking these changing features in the flow field, such as vortices and shock waves, researchers can have a deeper understanding of fluid dynamics phenomena. This is of great significance for revealing the evolution and interaction of flow field structures such as turbulence, vortices, and boundary layers. Feature tracking helps researchers define the occurrence of feature events by matching feature elements between adjacent time steps and analyzing the evolution of flow field features, such as continuation, splitting, and merging. Event detection is to describe the development and changes of features. Therefore, in order to explore and analyze the evolution of flow fields, feature tracking and its result visualization are important tasks in flow field scientific visualization. The existing classic attribute vector algorithm and volume overlap algorithm can solve the feature matching problem to a certain extent. However, the time-varying flow field data is huge. When using similarity metrics to match features, the classic algorithm needs to rely on the user to set the threshold for each attribute to make judgments, requiring users to have rich domain prior knowledge. Among them, the isosurface method is the most commonly used technology for visualizing the surface of the threshold area. However, most computer vision methods are aimed at two-dimensional image data. For three-dimensional time-varying flow fields, due to the large amount of data, complex data structure and viewpoint dependency, there are still many problems that need to be solved in terms of matching efficiency and interactive visualization. Therefore, it is necessary to design a more easily implemented feature tracking method to help researchers in need to analyze the behavior of features in the spatiotemporal domain.
发明内容Summary of the invention
有鉴于此，本发明的目的在于提供基于物理信息融合的流场特征提取跟踪方法、装置、设备，能够实现面向三维时序流场的涡特征准确提取和跟踪。其具体方案如下：In view of this, the purpose of the present invention is to provide a flow field feature extraction and tracking method, device, and equipment based on physical information fusion, which can realize accurate extraction and tracking of vortex features for three-dimensional time-series flow fields. The specific scheme is as follows:
第一方面，本申请公开了一种基于物理信息融合的流场特征提取跟踪方法，包括：In a first aspect, the present application discloses a flow field feature extraction and tracking method based on physical information fusion, comprising:
获取三维时序流场的各时间步下的流场数据；Obtain flow field data at each time step of the three-dimensional time series flow field;
识别各时间步下的所述流场数据中属于涡区域的数据点，并基于所有所述数据点对各时间步下的流场区域进行空间联通区域分割，以得到单时间步下的涡区域；Identify data points belonging to the vortex region in the flow field data at each time step, and perform spatially connected region segmentation on the flow field region at each time step based on all the data points to obtain the vortex region at a single time step;
计算各所述涡区域的各数据点对应的涡旋强度、涡度值、局部剪切率，以得到各所述涡区域的物理属性信息；Calculating the vortex intensity, vorticity value, and local shear rate corresponding to each data point in each vortex region to obtain physical property information of each vortex region;
基于各所述涡区域的物理属性信息分别构建所述涡区域的属性矩阵，并利用预设主成分分析法对各所述属性矩阵进行矩阵提取，以分别得到基于各所述涡区域的目标属性矩阵；Based on the physical property information of each vortex region, the attribute matrix of each vortex region is constructed respectively, and the matrix extraction of each attribute matrix is performed using a preset principal component analysis method to obtain a target attribute matrix based on each vortex region respectively;
根据相邻时间步下的所述目标属性矩阵之间的相似度计算结果构建特征相似度矩阵，根据预设特征对应关系遍历所述特征相似度矩阵，以判断所述三维时序流场在不同时间步之间变化中所属的目标流场事件类型。A feature similarity matrix is constructed according to the similarity calculation results between the target attribute matrices at adjacent time steps, and the feature similarity matrix is traversed according to the preset feature correspondence relationship to determine the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps.
可选的，所述识别各时间步下的所述流场数据中属于涡区域的数据点，并基于所有所述数据点对各时间步下的流场区域进行空间联通区域分割，以得到单时间步下的涡区域，包括：Optionally, the identifying data points belonging to the vortex region in the flow field data at each time step, and performing spatially connected region segmentation on the flow field region at each time step based on all the data points to obtain the vortex region at a single time step includes:
通过涡旋强度计算公式对各时间步下的所述流场数据进行涡旋强度计算，以分别得到各时间步下的所述流场数据中各流场数据点的涡旋强度；The vortex intensity of the flow field data at each time step is calculated by using a vortex intensity calculation formula to obtain the vortex intensity of each flow field data point in the flow field data at each time step;
将涡旋强度大于预设涡旋强度阈值的流场数据点作为涡区域的目标数据点；The flow field data points with vortex intensity greater than a preset vortex intensity threshold are taken as target data points of the vortex region;
选择所述涡区域中的任一目标数据点作为种子点，以所述种子点为起点，通过宽度优先搜索遍历单时间步下的所述目标数据点，为当前单时间步的特征区域创建一个原始队列，使用宽度优先搜索来遍历所述原始队列内的目标数据点及其周围邻域，且将满足预设区域阈值条件的目标数据点标记为区域特征，并将所述区域特征压进新队列，直至所述原始队列的目标数据点为空时停止，以得到单时间步下的涡区域。Select any target data point in the vortex area as a seed point, take the seed point as the starting point, traverse the target data points in a single time step through breadth-first search, create an original queue for the feature area of the current single time step, use breadth-first search to traverse the target data points and their surrounding neighborhoods in the original queue, mark the target data points that meet the preset regional threshold conditions as regional features, and push the regional features into the new queue until the target data points in the original queue are empty, so as to obtain the vortex area in a single time step.
可选的，所述计算各所述涡区域的各数据点对应的涡旋强度、涡度值、局部剪切率，以得到各所述涡区域的物理属性信息，包括：Optionally, the calculating of the vortex intensity, vorticity value, and local shear rate corresponding to each data point in each vortex region to obtain physical property information of each vortex region includes:
根据各所述涡区域的所述目标数据点的速度向量分量在不同速度方向的偏导数计算并确定所述目标数据点的涡度值；Calculate and determine the vorticity value of the target data point according to the partial derivatives of the velocity vector components of the target data point in each of the vortex regions in different velocity directions;
基于各所述涡区域的所述目标数据点的三维空间坐标信息、速度向量计算并确定所述目标数据点的雅各比矩阵；Calculate and determine the Jacobian matrix of the target data point based on the three-dimensional spatial coordinate information and velocity vector of the target data point in each vortex region;
根据所述雅各比矩阵以及所述雅各比矩阵的转置矩阵计算并确定所述目标数据点的张力张量矩阵；Calculate and determine the tension tensor matrix of the target data point according to the Jacobian matrix and the transposed matrix of the Jacobian matrix;
基于所述张力张量矩阵、所述涡度值计算并确定所述目标数据点的涡旋强度；Calculate and determine the vortex intensity of the target data point based on the tension tensor matrix and the vorticity value;
将所述涡旋强度、所述涡度值、所述雅各比矩阵的F范数、所述张力张量矩阵的F范数作为各所述涡区域的物理属性信息。The vortex intensity, the vorticity value, the F norm of the Jacobi matrix, and the F norm of the tension tensor matrix are used as physical property information of each vortex region.
可选的，所述基于各所述涡区域的物理属性信息分别构建所述涡区域的属性矩阵，并利用预设主成分分析法对各所述属性矩阵进行矩阵提取，以分别得到各所述涡区域的目标属性矩阵，包括：Optionally, the attribute matrices of the vortex regions are respectively constructed based on the physical attribute information of each vortex region, and matrix extraction is performed on each attribute matrix using a preset principal component analysis method to obtain a target attribute matrix of each vortex region, including:
按照所述目标数据点的所述涡旋强度、所述涡度值、所述雅各比矩阵的F范数、所述张力张量矩阵的F范数、三维空间坐标信息的顺序构建所述目标数据点的属性向量；Constructing the attribute vector of the target data point in the order of the vortex intensity, the vorticity value, the F norm of the Jacobi matrix, the F norm of the tension tensor matrix, and the three-dimensional space coordinate information of the target data point;
基于位于同一所述涡区域的目标数据点的属性向量构建的特征属性矩阵作为所述涡区域的属性矩阵，以得到各所述涡区域的属性矩阵；A characteristic attribute matrix constructed based on the attribute vectors of the target data points located in the same vortex region is used as the attribute matrix of the vortex region to obtain the attribute matrix of each vortex region;
对所述属性矩阵中的属性元素进行标准化处理，以得到标准化属性矩阵；Standardizing the attribute elements in the attribute matrix to obtain a standardized attribute matrix;
根据所述标准化属性矩阵计算对应的协方差矩阵；Calculate the corresponding covariance matrix according to the standardized attribute matrix;
利用协方差矩阵构建对应的投影矩阵，以根据所述投影矩阵和所述标准化属性矩阵确定各所述涡区域的目标属性矩阵。The corresponding projection matrix is constructed using the covariance matrix, so as to determine the target attribute matrix of each vortex region according to the projection matrix and the standardized attribute matrix.
可选的，所述利用协方差矩阵构建对应的投影矩阵，以根据所述投影矩阵和所述标准化属性矩阵确定各所述涡区域的目标属性矩阵，包括：Optionally, the using the covariance matrix to construct a corresponding projection matrix to determine the target attribute matrix of each vortex region according to the projection matrix and the standardized attribute matrix includes:
对所述协方差矩阵进行特征值分解，以计算所述协方差矩阵的特征值和相应的特征向量，并从所述特征值中筛选目标特征值和相应的目标特征向量构建投影矩阵；Performing eigenvalue decomposition on the covariance matrix to calculate eigenvalues and corresponding eigenvectors of the covariance matrix, and selecting target eigenvalues and corresponding target eigenvectors from the eigenvalues to construct a projection matrix;
利用所述投影矩阵对所述标准化属性矩阵进行投影降维，以得到相应的主成分属性矩阵，并将各所述主成分属性矩阵作为各所述涡区域的目标属性矩阵。The projection matrix is used to perform projection dimension reduction on the standardized attribute matrix to obtain a corresponding principal component attribute matrix, and each principal component attribute matrix is used as a target attribute matrix of each vortex region.
可选的，所述根据相邻时间步下的所述目标属性矩阵之间的相似度计算结果构建特征相似度矩阵之前，还包括：Optionally, before constructing the feature similarity matrix according to the similarity calculation results between the target attribute matrices at adjacent time steps, the method further includes:
计算相邻时间步下的所述目标属性矩阵之间的余弦相似度；Calculating the cosine similarity between the target attribute matrices at adjacent time steps;
对所述余弦相似度取平均值，以得到相邻时间步下的所述目标属性矩阵之间的相似度计算结果。The cosine similarities are averaged to obtain similarity calculation results between the target attribute matrices at adjacent time steps.
可选的，所述根据相邻时间步下的所述目标属性矩阵之间的相似度计算结果构建特征相似度矩阵，包括：Optionally, constructing a feature similarity matrix according to similarity calculation results between the target attribute matrices at adjacent time steps includes:
判断所述相似度计算结果与预设最小相似阈值之间的大小关系，若所述相似度计算结果小于所述预设最小相似阈值，则表征当前相邻时间步的特征区域不相似，并将当前相似度计算结果赋值0，以得到相应的第一赋值结果，然后将所述第一赋值结果作为构建特征相似度矩阵的矩阵元素；Determine the size relationship between the similarity calculation result and a preset minimum similarity threshold value. If the similarity calculation result is less than the preset minimum similarity threshold value, it indicates that the feature regions of the current adjacent time steps are not similar. The current similarity calculation result is assigned 0 to obtain a corresponding first assignment result. The first assignment result is then used as a matrix element for constructing a feature similarity matrix.
若所述相似度计算结果大于或等于所述预设最小相似阈值，则表征当前相邻时间步的特征区域相似，并将当前相似度计算结果赋值1，以得到相应的第二赋值结果，然后将所述第二赋值结果作为构建特征相似度矩阵的矩阵元素；If the similarity calculation result is greater than or equal to the preset minimum similarity threshold, it indicates that the feature regions of the current adjacent time steps are similar, and the current similarity calculation result is assigned a value of 1 to obtain a corresponding second assignment result, and then the second assignment result is used as a matrix element for constructing a feature similarity matrix;
利用所有所述相似度计算结果赋值后的赋值结果构建所述特征相似度矩阵。The feature similarity matrix is constructed using the assigned results of all the similarity calculation results.
可选的，所述根据预设特征对应关系遍历所述特征相似度矩阵，以判断所述三维时序流场在不同时间步之间变化中所属的目标流场事件类型，包括：Optionally, traversing the feature similarity matrix according to the preset feature correspondence relationship to determine the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps includes:
构建不同流场事件类型与流场特征之间的对应关系，以得到包含多个目标特征对应关系的预设特征对应关系；其中，所述流场事件类型包括：延续事件类型、分裂事件类型、合并事件类型、产生事件类型、耗散事件类型；Constructing correspondences between different flow field event types and flow field features to obtain a preset feature correspondence that includes multiple target feature correspondences; wherein the flow field event types include: continuation event type, split event type, merge event type, generation event type, and dissipation event type;
遍历所述特征相似度矩阵，以根据矩阵遍历结果以及预设特征对应关系确定目标特征对应关系；Traversing the feature similarity matrix to determine the target feature correspondence relationship according to the matrix traversal result and the preset feature correspondence relationship;
基于所述目标特征对应关系判断三维时序流场在不同时间步之间变化中所属的目标流场事件类型。Based on the target feature correspondence, the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps is determined.
第二方面，本申请公开了一种基于物理信息融合的流场特征提取跟踪装置，包括：In a second aspect, the present application discloses a flow field feature extraction and tracking device based on physical information fusion, comprising:
数据获取模块，用于获取三维时序流场的各时间步下的流场数据；A data acquisition module, used to acquire flow field data at each time step of a three-dimensional time series flow field;
区域分割模块，用于识别各时间步下的所述流场数据中属于涡区域的数据点，并基于所有所述数据点对各时间步下的流场区域进行空间联通区域分割，以得到单时间步下的涡区域；A region segmentation module is used to identify data points belonging to the vortex region in the flow field data at each time step, and to perform spatially connected region segmentation on the flow field region at each time step based on all the data points to obtain the vortex region at a single time step;
信息计算模块，用于计算各所述涡区域的各数据点对应的涡旋强度、涡度值、局部剪切率，以得到各所述涡区域的物理属性信息；An information calculation module, used for calculating the vortex intensity, vorticity value, and local shear rate corresponding to each data point of each vortex region, so as to obtain the physical property information of each vortex region;
矩阵构建模块，用于基于各所述涡区域的物理属性信息分别构建所述涡区域的属性矩阵，并利用预设主成分分析法对各所述属性矩阵进行矩阵提取，以分别得到各所述涡区域的目标属性矩阵；A matrix construction module, used for constructing attribute matrices of the vortex regions based on the physical attribute information of the vortex regions, and performing matrix extraction on the attribute matrices using a preset principal component analysis method to obtain target attribute matrices of the vortex regions;
跟踪模块，用于根据相邻时间步下的所述目标属性矩阵之间的相似度计算结果构建特征相似度矩阵，根据预设特征对应关系遍历所述特征相似度矩阵，以判断所述三维时序流场在不同时间步之间变化中所属的目标流场事件类型。The tracking module is used to construct a feature similarity matrix based on the similarity calculation results between the target attribute matrices at adjacent time steps, and traverse the feature similarity matrix according to the preset feature correspondence relationship to determine the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps.
第三方面，本申请公开了一种电子设备，包括：In a third aspect, the present application discloses an electronic device, comprising:
存储器，用于保存计算机程序；Memory, used to store computer programs;
处理器，用于执行所述计算机程序，以实现前述公开的基于物理信息融合的流场特征提取跟踪方法的步骤。The processor is used to execute the computer program to implement the steps of the flow field feature extraction and tracking method based on physical information fusion disclosed above.
第四方面，本申请公开了一种计算机可读存储介质，用于存储计算机程序；其中，所述计算机程序被处理器执行时实现前述公开的基于物理信息融合的流场特征提取跟踪方法的步骤。In a fourth aspect, the present application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the aforementioned disclosed flow field feature extraction and tracking method based on physical information fusion are implemented.
可见，本申请公开了获取三维时序流场的各时间步下的流场数据；识别各时间步下的所述流场数据中属于涡区域的数据点，并基于所有所述数据点对各时间步下的流场区域进行空间联通区域分割，以得到单时间步下的涡区域；计算各所述涡区域的各数据点对应的涡旋强度、涡度值、局部剪切率，以得到各所述涡区域的物理属性信息；基于各所述涡区域的物理属性信息分别构建所述涡区域的属性矩阵，并利用预设主成分分析法对各所述属性矩阵进行矩阵提取，以分别得到各所述涡区域的目标属性矩阵；根据相邻时间步下的所述目标属性矩阵之间的相似度计算结果构建特征相似度矩阵，根据预设特征对应关系遍历所述特征相似度矩阵，以判断所述三维时序流场在不同时间步之间变化中所属的目标流场事件类型。由此可见，通过对于各时间步下的流场区域进行区域分割，进而获取到单个时间步下的涡区域，以便根据涡区域中的流场数据计算其物理属性信息，然后对物理属性信息进行融合简化，获取目标属性矩阵，最后通过相邻时间步下的目标属性矩阵的相似度结果进一步构建特征相似度矩阵，对特征相似度矩阵与预设特征对应关系进行匹配，实现三维流场涡特征间的匹配和事件跟踪，减少了阈值依赖，提高了计算效率。It can be seen that the present application discloses the acquisition of flow field data at each time step of a three-dimensional time series flow field; identification of data points belonging to the vortex region in the flow field data at each time step, and spatially connected region segmentation of the flow field region at each time step based on all the data points to obtain the vortex region at a single time step; calculation of the vortex intensity, vorticity value, and local shear rate corresponding to each data point in each vortex region to obtain physical property information of each vortex region; construction of attribute matrices of the vortex region based on the physical property information of each vortex region, and matrix extraction of each attribute matrix using a preset principal component analysis method to obtain target attribute matrices of each vortex region; construction of a feature similarity matrix based on the similarity calculation results between the target attribute matrices at adjacent time steps, and traversal of the feature similarity matrix according to the preset feature correspondence to determine the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps. It can be seen that by segmenting the flow field area at each time step, the vortex area at a single time step is obtained, so that the physical property information can be calculated according to the flow field data in the vortex area, and then the physical property information is fused and simplified to obtain the target attribute matrix. Finally, the feature similarity matrix is further constructed through the similarity results of the target attribute matrix at adjacent time steps, and the feature similarity matrix is matched with the preset feature correspondence to achieve matching and event tracking between three-dimensional flow field vortex features, reduce threshold dependence, and improve computational efficiency.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据提供的附图获得其他的附图。In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
图1为本申请公开的一种基于物理信息融合的流场特征提取跟踪方法流程图；FIG1 is a flow chart of a flow field feature extraction and tracking method based on physical information fusion disclosed in the present application;
图2为本申请公开的当T=54时的涡区域特征分割结果图；FIG2 is a diagram showing the vortex region feature segmentation result when T=54 disclosed in the present application;
图3为本申请公开的当T=55时的涡区域特征分割结果图；FIG3 is a diagram showing the vortex region feature segmentation result when T=55 disclosed in the present application;
图4为本申请公开的一种涡区域增长图；FIG4 is a vortex region growth diagram disclosed in the present application;
图5为本申请公开的一种具体的基于物理信息融合的流场特征提取跟踪方法流程图；FIG5 is a flow chart of a specific flow field feature extraction and tracking method based on physical information fusion disclosed in the present application;
图6为本申请公开的当T=54与T=55时的特征相似矩阵；FIG6 is a feature similarity matrix when T=54 and T=55 disclosed in this application;
图7为本申请公开的一种不同类型事件与特征对应关系示意图；FIG7 is a schematic diagram of the correspondence between different types of events and features disclosed in the present application;
图8为本申请公开的一种T时刻和T+1时刻的流场特征提取跟踪方法流程图；FIG8 is a flow chart of a flow field feature extraction and tracking method at time T and time T+1 disclosed in the present application;
图9为本申请公开的一种基于物理信息融合的流场特征提取跟踪装置结构示意图；FIG9 is a schematic diagram of the structure of a flow field feature extraction and tracking device based on physical information fusion disclosed in the present application;
图10为本申请公开的一种电子设备结构图。FIG. 10 is a structural diagram of an electronic device disclosed in this application.
具体实施方式DETAILED DESCRIPTION
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
科学可视化是研究如何运用计算机图形学和图像处理技术，将科学计算产生的数据及科学测量获得的数据，转换为图形或图像在屏幕上显示出来，并进行交互处理的理论、方法和技术的领域。流场可视化是科学可视化的一个重要研究方向，是流场科学计算，即计算流体力学（Computational Fluid Dynamics，CFD）不可或缺的重要组成部分。流场可视化在分析和理解复杂的流场流动机理，洞察流场物理现象和发现流动科学规律方面，正在发挥越来越重要的作用。二十多年来，流场可视化是科学计算可视化领域在技术方法研究和应用软件工具开发中取得成果最多的方向之一。随着科技的发展，工业部门对流场仿真的认识越来越深刻，建立的流场模型也越来越贴近真实物理环境，对流场可视化的精细程度和精度提出了更高的要求，进而导致了流场仿真数据的数据量不断攀升，甚至能达到TB甚至PB量级。对海量数据进行存储和读取，势必引起极大的开销，数据I/O开销已经逐渐成为了CFD计算的瓶颈之一。为了解决海量数据I/O难题，一种新型可视化方法“原位可视化”应运而生。其优势在于无需将流场解算结果写入磁盘，而是在解算的同时将用户所需的结果直接可视化呈现，可视化管线直接从程序内存中获取数据，减少了总开销。原位可视化中一个关键的问题是在进行可视化时如何对用户关心的流产关键信息进行特征提取，并对特征随着时间的变化进行事件跟踪。但是，流场科学计算数据类型复杂、物理量种类多、物理特征提取困难、可视化要求很高，使得特征跟踪一直是流场可视化领域的研究热点和难点之一。Scientific visualization is a field that studies how to use computer graphics and image processing technology to convert data generated by scientific computing and data obtained by scientific measurement into graphics or images to display on the screen and perform interactive processing. Flow field visualization is an important research direction of scientific visualization and an indispensable part of flow field scientific computing, namely computational fluid dynamics (CFD). Flow field visualization is playing an increasingly important role in analyzing and understanding the flow mechanism of complex flow fields, gaining insight into flow field physical phenomena and discovering the laws of flow science. For more than 20 years, flow field visualization has been one of the directions with the most achievements in the research of technical methods and the development of application software tools in the field of scientific computing visualization. With the development of science and technology, the industrial sector has a deeper understanding of flow field simulation, and the established flow field models are getting closer and closer to the real physical environment, which puts forward higher requirements for the refinement and accuracy of flow field visualization, which in turn leads to the continuous increase in the amount of flow field simulation data, which can even reach the TB or PB level. Storing and reading massive data will inevitably cause huge overhead, and data I/O overhead has gradually become one of the bottlenecks of CFD calculations. In order to solve the problem of massive data I/O, a new visualization method "in-situ visualization" came into being. Its advantage is that there is no need to write the flow field solution results to the disk, but to directly visualize the results required by the user while solving the flow field. The visualization pipeline directly obtains data from the program memory, reducing the total overhead. A key issue in in-situ visualization is how to extract features of the key flow information that users are concerned about during visualization, and track events as the features change over time. However, the data types of flow field scientific calculations are complex, the types of physical quantities are many, the physical features are difficult to extract, and the visualization requirements are very high, making feature tracking one of the research hotspots and difficulties in the field of flow field visualization.
特征提取和跟踪是一种成熟的技术，用于分析各种领域的时变数据，如视频分析、计算机视觉和流场可视化。在这些领域中，特征跟踪都是通过首先提取特征，然后随时间跟踪来应用的。为了跟踪单个特征的演化，必须在不同的时间步对特征进行一致分类。否则，特征信息无法在可视化中连贯地呈现。例如，如果使用颜色来区分不同的特征，那么单个特征必须始终保持相同的颜色，这样颜色信息才有帮助。这个任务通常被称为对应问题。现有的特征提取技术主要分为两大类：基于对应关系的方法和基于时空的方法。目前流行的大多数特征跟踪技术都是基于对应的方法，分别每个时间步所提取的特征，然后根据位置、大小、方向或区域重叠等标准将提取的特征在连续的时间步中进行对应。时空提取方法将时间轴视为第四维空间轴，然后将区域生长等标准特征提取技术一次性应用于整个时空域。提取结果因方法而异，并且通常依赖于数据。许多应用使用基于图像处理技术的方法，如区域生长、阈值分割和边缘检测，这些方法通常用于医疗数据进行数据分割。Feature extraction and tracking is a well-established technique for analyzing time-varying data in a variety of fields, such as video analysis, computer vision, and flow visualization. In each of these fields, feature tracking is applied by first extracting features and then tracking them over time. In order to track the evolution of a single feature, the features must be consistently classified at different time steps. Otherwise, the feature information cannot be presented coherently in the visualization. For example, if color is used to distinguish different features, a single feature must always remain the same color so that the color information is helpful. This task is often referred to as the correspondence problem. Existing feature extraction techniques can be divided into two main categories: correspondence-based methods and spatiotemporal-based methods. Most of the popular feature tracking techniques are based on correspondence methods, which extract features at each time step and then correspond the extracted features at consecutive time steps based on criteria such as position, size, orientation, or area overlap. Spatiotemporal extraction methods treat the time axis as the fourth spatial dimension and then apply standard feature extraction techniques such as region growing to the entire spatiotemporal domain at once. The extraction results vary from method to method and are often data-dependent. Many applications use methods based on image processing techniques such as region growing, threshold segmentation, and edge detection, which are often used for data segmentation in medical data.
流场可视化领域的特征跟踪与计算机视觉领域的相关性问题类似。许多在计算机视觉领域的技术和方法可以被用来提取、辨识并跟踪特征。流场可视化领域存在一些常用的标量提取方法。常用的用于提取感兴趣区域的方法是阈值法，即根据给定阈值的大小来判定特征区域的范围。等值面方法是用于可视化阈值区域表面的最常用的技术。但计算机视觉领域的方法大多针对二维图像数据，对于三维时变流场而言，由于其数据量大、数据结构复杂且与视点相关，在匹配效率、交互可视化等方面仍存在较多问题亟需解决。通过跟踪流场中这些特征的变化，如涡、激波等，研究者可以更深入地理解流体动力学现象。这对于揭示湍流、涡旋、边界层等流场结构的演变和相互作用有重要意义。特征跟踪通过在相邻时间步长之间匹配特征元素，分析流场特征的演化历程，从而帮助研究人员定义特征事件的发生，例如，延续、分裂、合并等，事件检测就是为了描述特征所发生的发展和变化。因此，为了探索和分析流场的演化过程，要对流场进行特征跟踪，将相邻时刻特征间的关系进行匹配。Feature tracking in the field of flow field visualization is similar to the correlation problem in the field of computer vision. Many technologies and methods in the field of computer vision can be used to extract, identify and track features. There are some commonly used scalar extraction methods in the field of flow field visualization. The commonly used method for extracting the region of interest is the threshold method, that is, the range of the feature area is determined according to the size of the given threshold. The isosurface method is the most commonly used technique for visualizing the surface of the threshold area. However, most of the methods in the field of computer vision are aimed at two-dimensional image data. For three-dimensional time-varying flow fields, due to the large amount of data, complex data structure and viewpoint dependence, there are still many problems in matching efficiency and interactive visualization that need to be solved. By tracking the changes of these features in the flow field, such as vortices and shock waves, researchers can have a deeper understanding of fluid dynamics phenomena. This is of great significance for revealing the evolution and interaction of flow field structures such as turbulence, vortices, and boundary layers. Feature tracking helps researchers define the occurrence of feature events by matching feature elements between adjacent time steps and analyzing the evolution of flow field features, such as continuation, splitting, merging, etc. Event detection is to describe the development and changes of features. Therefore, in order to explore and analyze the evolution of the flow field, it is necessary to track the characteristics of the flow field and match the relationship between the characteristics at adjacent moments.
针对计算流体动力学（CFD）中的涡旋跟踪需求，现有技术1利用深度优先搜索（Depth-First Search，DFS）算法标记和跟踪涡旋，用于跟踪和观察计算流体动力学中的涡旋结构。该算法使用Q准则来识别涡旋，通过渲染给定Q阈值的等值面来可视化涡旋。将计算网格视为一个环状图，流场单元中心是图的节点，单元之间的邻接连接由图的边表示。求得相邻时间涡旋交集后，将前一时间步涡旋单元边界作为起点，使用深度优先搜索（DFS）算法来遍历图，标记涡旋区域。使用洪水填充（Flood Fill）算法为每个涡旋分配一个唯一的标识符。通过比较连续时间步长中的涡旋场，为每个涡旋分配索引，以确保它们在时间上是连续的。但是现有技术1适用于分析和理解二维情况下湍流中的相干结构，能够量化涡旋之间的相互作用，如合并和撕裂事件，但并不适用于三维流场中的特征跟踪。现有技术2提出了一个基于预测和校正的特征跟踪方法，其首先预测每个特征的位置，然后调整预测后的表面矫正达到正确的位置。此方法使用了前序时间步的特征信息来预测当前时间步特征的位置，并通过区域增长和缩减算法调整预测后特征的表面。结合了特征的动态性和一致性，适用于科学计算中的大规模数据集。但同时方法的性能依赖于预测的准确性，可能需要更多的工作来校正预测区域。提出的预测函数主要基于对象的运动，没有考虑其他可能影响预测准确性的特征，如区域合并和分裂时的不连续性。现有技术3提出了一个跟踪三维时变流场管状涡特征的算法。管状涡是一个被”骨骼”表示描述的长条结构。首先，一个域中的种子点被交互的选取，并且一个管状探头被用来计算在管内所有涡量向量的平均长度。这个过程在由种子点获得的管状体中不断重复，而具有最大平均值的管状体则成为了一个管状涡的分割。当最佳分割线被找到时，种子点便被移到分割的另一个端点以便后续搜索，直到满足约束。逐渐创建的骨骼被用来在下一个时间点跟踪管状涡。根据新结构会大约平行于原来的结构以及管状涡漂流的距离较短的经验，用螺旋搜索技术来确定管状涡的位置，但是仅适用于管状涡结构的跟踪。现有技术4基于特征提取结果，提出了一种基于空间分布优化的特征相似度度量方法，该方法使用特征体积的重叠度结合特征的空间分布信息求特征间的相似度。首先基于点之间的欧几里得距离计算连续时间内两个不同特征点集的最小移动代价，再计算两个特征之间的空间重叠度，通过将最小移动代价与空间重叠度加权相加最终求得两个特征之间的相似度。虽然通过全局优化避免了对阈值的依赖，减少了因阈值选择不当引入的误差。但全局优化和空间分布相似性仅根据特征的空间属性进行跟踪而忽略了特征具有的流体物理属性。In response to the vortex tracking needs in computational fluid dynamics (CFD), prior art 1 uses a depth-first search (DFS) algorithm to mark and track vortices for tracking and observing vortex structures in computational fluid dynamics. The algorithm uses the Q criterion to identify vortices and visualizes vortices by rendering isosurfaces with a given Q threshold. The computational grid is regarded as a ring graph, the center of the flow field unit is the node of the graph, and the adjacency connection between units is represented by the edge of the graph. After obtaining the intersection of adjacent time vortices, the boundary of the vortex unit in the previous time step is used as the starting point, and the depth-first search (DFS) algorithm is used to traverse the graph and mark the vortex area. A unique identifier is assigned to each vortex using the Flood Fill algorithm. By comparing the vortex fields in consecutive time steps, an index is assigned to each vortex to ensure that they are continuous in time. However, prior art 1 is suitable for analyzing and understanding coherent structures in turbulence in two-dimensional situations, and can quantify the interactions between vortices, such as merging and tearing events, but is not suitable for feature tracking in three-dimensional flow fields. Prior art 2 proposes a feature tracking method based on prediction and correction, which first predicts the position of each feature and then adjusts the predicted surface correction to the correct position. This method uses the feature information of the previous time step to predict the position of the feature in the current time step, and adjusts the surface of the predicted feature through a region growth and reduction algorithm. It combines the dynamics and consistency of the feature and is suitable for large-scale data sets in scientific computing. But at the same time, the performance of the method depends on the accuracy of the prediction, and more work may be required to correct the predicted region. The proposed prediction function is mainly based on the movement of the object, and does not consider other features that may affect the accuracy of the prediction, such as discontinuities during region merging and splitting. Prior art 3 proposes an algorithm for tracking tubular vortex features in a three-dimensional time-varying flow field. A tubular vortex is a long strip structure described by a "skeleton" representation. First, seed points in a domain are interactively selected, and a tubular probe is used to calculate the average length of all vortex vectors in the tube. This process is repeated in the tubular body obtained by the seed point, and the tubular body with the largest average value becomes a segmentation of the tubular vortex. When the best segmentation line is found, the seed point is moved to the other end point of the segmentation for subsequent search until the constraint is met. The gradually created skeleton is used to track the tubular vortex at the next time point. Based on the experience that the new structure will be approximately parallel to the original structure and the tubular vortex drifts a short distance, the spiral search technique is used to determine the position of the tubular vortex, but it is only applicable to the tracking of tubular vortex structures. Based on the feature extraction results, the prior art 4 proposes a feature similarity measurement method based on spatial distribution optimization. This method uses the overlap of the feature volume combined with the spatial distribution information of the feature to calculate the similarity between features. First, the minimum moving cost of two different feature point sets in continuous time is calculated based on the Euclidean distance between the points, and then the spatial overlap between the two features is calculated. The similarity between the two features is finally obtained by weighted addition of the minimum moving cost and the spatial overlap. Although the dependence on the threshold is avoided by global optimization, the error introduced by improper threshold selection is reduced. However, global optimization and spatial distribution similarity only track according to the spatial properties of the features and ignore the fluid physical properties of the features.
为此，本发明提供了一种基于物理信息融合的流场特征提取跟踪方法，能够实现面向三维时序流场的涡特征准确提取和跟踪。To this end, the present invention provides a flow field feature extraction and tracking method based on physical information fusion, which can realize accurate extraction and tracking of vortex features for three-dimensional time-series flow fields.
参照图1所示，本发明实施例公开了一种基于物理信息融合的流场特征提取跟踪方法，包括：1 , an embodiment of the present invention discloses a flow field feature extraction and tracking method based on physical information fusion, comprising:
步骤S11：获取三维时序流场的各时间步下的流场数据。Step S11: Obtain flow field data at each time step of the three-dimensional time series flow field.
本实施例中，获取三维时序流场在各时间步下的速度分量（x，y，z三个方向）、流场各点的压力分布信息、流体密度、涡量信息等流场数据，需要注意的是，本发明的流场数据是各个时间步下按照时间顺序获取的，不同时间步的流场数据能反映出流场随时间的变化过程和趋势，从而深入研究其动态特性，而且一些关键的涡特征只在特定时间出现或发生变化，获取各时间步数据有助于捕捉这些短暂而重要的现象，最后，为了实现本发明中的流场特征（涡特征）在时间序列上的准确跟踪，由于流场不是静态的，其特征随时间不断变化，各时间步数据能完整呈现这种演化过程，以便对比不同时间点的流场状态，进而用于发现规律和异常需要各时间步的流场数据作为基础。In this embodiment, flow field data such as velocity components (x, y, and z directions) of the three-dimensional time-series flow field at each time step, pressure distribution information at each point in the flow field, fluid density, vortex information, etc. are obtained. It should be noted that the flow field data of the present invention is obtained in chronological order at each time step. The flow field data at different time steps can reflect the change process and trend of the flow field over time, so as to deeply study its dynamic characteristics. Moreover, some key vortex features only appear or change at a specific time. Acquiring data at each time step helps to capture these short-lived and important phenomena. Finally, in order to achieve accurate tracking of the flow field characteristics (vortex characteristics) in the time series in the present invention, since the flow field is not static and its characteristics are constantly changing with time, the data at each time step can fully present this evolution process, so as to compare the flow field states at different time points, and then use the flow field data at each time step as a basis for discovering laws and anomalies.
步骤S12：识别各时间步下的所述流场数据中属于涡区域的数据点，并基于所有所述数据点对各时间步下的流场区域进行空间联通区域分割，以得到单时间步下的涡区域。Step S12: Identify the data points belonging to the vortex region in the flow field data at each time step, and perform spatially connected region segmentation on the flow field region at each time step based on all the data points to obtain the vortex region at a single time step.
本实施例中，通过涡旋强度计算公式对各时间步下的所述流场数据进行涡旋强度计算，以分别得到各时间步下的所述流场数据中各流场数据点的涡旋强度；可以理解的是，首先，对各时间步下的流场数据进行涡旋强度计算，以得到各时间步下的各流场数据点对应的涡旋强度，以便根据该涡旋强度进一步进行特征区域（涡区域）的识别划分，具体的，通过涡旋强度计算公式计算各流场数据点的涡旋强度，公式如下：In this embodiment, the vortex strength calculation formula is used to calculate the vortex strength of the flow field data at each time step to obtain the vortex strength of each flow field data point in the flow field data at each time step. It can be understood that, first, the vortex strength calculation is performed on the flow field data at each time step to obtain the vortex strength corresponding to each flow field data point at each time step, so as to further identify and divide the characteristic area (vortex area) according to the vortex strength. Specifically, the vortex strength of each flow field data point is calculated by the vortex strength calculation formula, and the formula is as follows:
； ;
其中，Q表示涡旋强度，Ω表示各流场数据点的涡度值，S表示各流场数据点的张力张量矩阵。|| ||表示Ω和S的范数。此外，对于各流场数据进行涡旋强度计算以便后续涡区域识别的步骤，还可以通过Ω方法、λ2方法来实现，对此不做具体限定。Wherein, Q represents the vortex intensity, Ω represents the vorticity value of each flow field data point, and S represents the tension tensor matrix of each flow field data point. || || represents the norm of Ω and S. In addition, the step of calculating the vortex intensity for each flow field data for subsequent vortex region identification can also be achieved by the Ω method and the λ2 method, which are not specifically limited.
本实施例中，将涡旋强度大于预设涡旋强度阈值的流场数据点作为涡区域的目标数据点；可以理解的是，当计算出各个流场数据点的涡旋强度后，通过判断涡旋强度Q是否大于预设涡旋强度阈值，确定对应的流场数据点是否涡区域的数据点，具体的，将预设涡旋强度阈值设置为0，当Q＞0时，将对应的流场数据点标记为涡区域的目标数据点，这样一来，可以将那些旋转运动较强、涡旋特征明显的数据点识别为涡区域的目标数据点，从而进行后续的分析和处理。然后，基于各时间步下的目标数据点对各个时间步的流场区域进行空间联通区域分割，得到单时间步下的涡区域。其中，选择预设涡旋强度阈值设置为0，将Q＞0作为筛选目标数据点的条件，是因为意味Q＞0表征涡度值的平方大于应变率的平方，即流体的旋转运动占据主导地位，符合涡旋的特征，因此通过Q＞0筛选出具有明显涡旋特征的目标数据点。需要注意的是，若直接将所有满足条件的目标数据点聚合为涡区域可能会包含噪声和孤立点，这些点可能并不真正属于涡区，因此需要进一步采用三维区域增长算法比较相邻目标数据点的特征，将相似的目标数据点连接起来，形成更大的连通区域（涡区域），从而去除噪声和孤立点的影响。In this embodiment, the flow field data points whose vortex intensity is greater than the preset vortex intensity threshold are used as the target data points of the vortex area; it can be understood that after calculating the vortex intensity of each flow field data point, by judging whether the vortex intensity Q is greater than the preset vortex intensity threshold, it is determined whether the corresponding flow field data point is a data point of the vortex area. Specifically, the preset vortex intensity threshold is set to 0. When Q>0, the corresponding flow field data point is marked as the target data point of the vortex area. In this way, those data points with strong rotational motion and obvious vortex characteristics can be identified as target data points of the vortex area, so as to perform subsequent analysis and processing. Then, based on the target data points at each time step, the flow field area of each time step is spatially connected to obtain the vortex area under a single time step. Among them, the preset vortex intensity threshold is set to 0, and Q>0 is used as the condition for screening the target data point, because it means that Q>0 indicates that the square of the vortex value is greater than the square of the strain rate, that is, the rotational motion of the fluid occupies a dominant position, which conforms to the characteristics of the vortex, so Q>0 is used to screen out the target data points with obvious vortex characteristics. It should be noted that if all target data points that meet the conditions are directly aggregated into vortex regions, they may contain noise and isolated points, which may not actually belong to the vortex region. Therefore, it is necessary to further use a three-dimensional region growing algorithm to compare the features of adjacent target data points and connect similar target data points to form a larger connected region (vortex region), thereby removing the influence of noise and isolated points.
进一步的，对目标数据点使用三维区域增长算法进行空间连通区域分割，得到单个时间步所有的涡区域，如图2所示为时间步T=54时的各涡区域特征分割结果图，包含涡区域特征0、涡区域特征1、涡区域特征2、涡区域特征3、涡区域特征4的涡区域位置信息，以及部分涡区域特征之间的连通信息，其中，在图2中的红色区域、橙色区域、黄色区域、青色区域、蓝色区域分别表示时间步T=54时涡区域中涡旋强度Q值的从大到小的顺序排布，其中，黑色部分为背景区域；如图3所示为时间步T=55时的各涡区域特征分割结果图，包含涡区域特征0、涡区域特征1、涡区域特征2、涡区域特征3、涡区域特征4的涡区域位置信息，以及部分涡区域特征之间的连通信息，其中，在图3中的红色区域、橙色区域、黄色区域、青色区域、蓝色区域分别表示时间步T=55时涡区域中涡旋强度Q值的从大到小的顺序排布，其中，黑色部分为背景区域；能够从图2和图3在相邻时间步下的各自涡区域特征分割结果图中包含的涡区域特征位置信息以及相互之间的连通信息等位置信息变化和连通信息变化，进一步获取变化的特征信息。其中，为了得到单个时间步所有的涡区域，采用三维区域增长算法并在三维体素数据中识别和分割具有相似特性的区域的特点，在三维空间的流场区域进行涡区域分割处理，具体的，选择所述涡区域中的任一目标数据点作为种子点，以所述种子点为起点，通过宽度优先搜索遍历单时间步下的所述目标数据点，为当前单时间步的特征区域创建一个原始队列，使用宽度优先搜索来遍历所述原始队列内的目标数据点及其周围邻域，且将满足预设区域阈值条件的目标数据点标记为区域特征，并将所述区域特征压进新队列，直至所述原始队列的目标数据点为空时停止，以得到单时间步下的涡区域。可以理解的是，需要将当前涡区域中的种子点作为起始点，如果没有指定种子点，则会自动将搜索到的任一目标数据点设定为种子点。获得种子点的位置后，以当前确定的种子点为起点，通过宽度优先搜索进行遍历，为当前时刻（当前单时间步）的特征区域（涡区域）创建一个原始队列，该原始队列的初始元素为当前种子点，随后，使用宽度优先搜索来遍历原始队列内的目标数据点及其周围邻域，且不断将满足区域阈值条件的目标数据点标记为区域特征并压进新队列，直至此原始队列内目标数据点为空时停止，此时算法所遍历过的存到新队列中的所有目标数据点便形成了该特征区域的空间分布。其中，区域阈值条件为判断目标数据点是否属于同一特征区域的准则。具体的阈值条件可以根据具体的应用和数据特点进行设定，其中，常见的区域阈值条件是基于目标数据点的数值属性，例如灰度值、强度值或其他特征值。通过设置合适的阈值，可以将目标数据点分为不同的特征区域。在实际应用中，阈值的选择通常需要根据经验、数据分布和具体需求进行调整。可以通过试验不同的阈值来找到最适合的分割结果，对此不作具体限定。此外，还可以结合其他条件或方法来进一步优化区域增长算法的性能，例如：使用种子点的选择策略、考虑邻域关系、采用多阈值或自适应阈值等。图4为本发明区域增长算法的二维示意图，其中红色线段所包围的区域为当前特征区域，蓝色区域和箭头分别为算法已经遍历的区域和算法遍历的方向。Furthermore, the three-dimensional region growing algorithm is used to segment the spatially connected regions of the target data points to obtain all the vortex regions in a single time step. As shown in Figure 2, the feature segmentation result diagram of each vortex region at time step T=54 includes the vortex region position information of vortex region feature 0, vortex region feature 1, vortex region feature 2, vortex region feature 3, and vortex region feature 4, as well as the connectivity information between some vortex region features. The red area, orange area, yellow area, cyan area, and blue area in Figure 2 respectively represent the order of the vortex intensity Q value in the vortex region at time step T=54 from large to small, and the black part is the background area; as shown in Figure 3, the vortex regions at time step T=55 are shown in Figure 4. The regional feature segmentation result map includes the vortex region position information of vortex region feature 0, vortex region feature 1, vortex region feature 2, vortex region feature 3, and vortex region feature 4, as well as the connectivity information between some vortex region features, wherein the red area, orange area, yellow area, cyan area, and blue area in FIG3 respectively represent the order of the vortex intensity Q value in the vortex region at time step T=55 from large to small, wherein the black part is the background area; the changed feature information can be further obtained from the changes in position information and connectivity information such as the vortex region feature position information and the connectivity information between each other contained in the respective vortex region feature segmentation result maps of FIG2 and FIG3 at adjacent time steps. Among them, in order to obtain all the vortex regions in a single time step, a three-dimensional region growth algorithm is used to identify and segment regions with similar characteristics in three-dimensional voxel data, and vortex region segmentation processing is performed in the flow field region of three-dimensional space. Specifically, any target data point in the vortex region is selected as a seed point, and the seed point is used as the starting point. The target data point under a single time step is traversed through a breadth-first search, and an original queue is created for the feature region of the current single time step. The target data point and its surrounding neighborhood in the original queue are traversed using a breadth-first search, and the target data point that meets the preset regional threshold condition is marked as a regional feature, and the regional feature is pressed into the new queue until the target data point in the original queue is empty, so as to obtain the vortex region under a single time step. It can be understood that the seed point in the current vortex region needs to be used as the starting point. If the seed point is not specified, any target data point searched will be automatically set as the seed point. After obtaining the position of the seed point, the currently determined seed point is used as the starting point, and a breadth-first search is used to traverse the feature area (vortex area) at the current moment (current single time step). An original queue is created, and the initial element of the original queue is the current seed point. Subsequently, the breadth-first search is used to traverse the target data points in the original queue and their surrounding neighborhoods, and the target data points that meet the regional threshold conditions are continuously marked as regional features and pressed into the new queue until the target data points in the original queue are empty. At this time, all the target data points stored in the new queue traversed by the algorithm form the spatial distribution of the feature area. Among them, the regional threshold condition is the criterion for judging whether the target data points belong to the same feature area. The specific threshold condition can be set according to the specific application and data characteristics. Among them, the common regional threshold condition is based on the numerical attributes of the target data points, such as grayscale value, intensity value or other feature values. By setting a suitable threshold, the target data points can be divided into different feature areas. In practical applications, the selection of thresholds usually needs to be adjusted according to experience, data distribution and specific needs. The most suitable segmentation result can be found by experimenting with different thresholds, and there is no specific limitation on this. In addition, other conditions or methods can be combined to further optimize the performance of the region growing algorithm, for example: using a seed point selection strategy, considering neighborhood relationships, using multiple thresholds or adaptive thresholds, etc. FIG4 is a two-dimensional schematic diagram of the region growing algorithm of the present invention, wherein the area surrounded by the red line segment is the current feature area, and the blue area and arrows are the area that the algorithm has traversed and the direction of the algorithm traversal, respectively.
步骤S13：计算各所述涡区域的各数据点对应的涡旋强度、涡度值、局部剪切率，以得到各所述涡区域的物理属性信息。Step S13: Calculate the vortex intensity, vorticity value, and local shear rate corresponding to each data point of each vortex region to obtain physical property information of each vortex region.
本实施例中，根据各所述涡区域的所述目标数据点的速度向量分量在不同速度方向的偏导数计算并确定所述目标数据点的涡度值；可以理解的是，In this embodiment, the vorticity value of the target data point is calculated and determined according to the partial derivatives of the velocity vector components of the target data point in each vortex region in different velocity directions; it can be understood that
按照计算并确定涡区域中目标数据点的涡度向量，其中，表示目标数据点的涡度值，分别表示涡度向量在方向上的分量，表示梯度运算符，×表示叉积运算符，V表示速度向量，表示目标数据点的三维空间坐标，表示在方向上的速度向量分量。然后通过计算涡度向量的模确定目标数据点的涡度值W。according to Calculate and determine the vorticity vector of the target data point in the vortex region, where represents the vorticity value of the target data point, They represent the vorticity vector The weight in direction, represents the gradient operator, × represents the cross product operator, V represents the velocity vector, Represents the three-dimensional spatial coordinates of the target data point, Indicated in The velocity vector component in the direction. Then the vorticity value W of the target data point is determined by calculating the modulus of the vorticity vector.
本实施例中，基于各所述涡区域的所述目标数据点的三维空间坐标信息、速度向量计算并确定所述目标数据点的雅各比矩阵；其中，雅各比矩阵如下：In this embodiment, the Jacobian matrix of the target data point is calculated and determined based on the three-dimensional spatial coordinate information and velocity vector of the target data point in each vortex region; wherein the Jacobian matrix is as follows:
； ;
其中，表示目标数据点的三维空间坐标信息，表示目标数据点的速度向量，通过雅各比矩阵的各个元素表示了速度向量在不同方向上的偏导数，它描述了流场的空间变化率，便于后续通过分析雅各比矩阵的特征值和特征向量，可以了解流场的稳定性、涡旋结构等重要信息。in, Represents the three-dimensional spatial coordinate information of the target data point, The velocity vector represents the target data point. The partial derivatives of the velocity vector in different directions are represented by the elements of the Jacobi matrix. It describes the spatial rate of change of the flow field, which facilitates the subsequent analysis of the eigenvalues and eigenvectors of the Jacobi matrix to understand important information such as the stability of the flow field and the vortex structure.
本实施例中，根据所述雅各比矩阵以及所述雅各比矩阵的转置矩阵计算并确定所述目标数据点的张力张量矩阵；可以理解的是，利用雅各比矩阵和转置矩阵进一步确定目标数据点的张力张量矩阵，表示如下：In this embodiment, the tension tensor matrix of the target data point is calculated and determined according to the Jacobian matrix and the transposed matrix of the Jacobian matrix; it can be understood that the tension tensor matrix of the target data point is further determined using the Jacobian matrix and the transposed matrix, which is expressed as follows:
； ;
其中，J表示雅各比矩阵，JT表示雅各比矩阵的转置矩阵，S表示张力张量矩阵。Among them, J represents the Jacobian matrix, JT represents the transposed matrix of the Jacobian matrix, and S represents the tension tensor matrix.
本实施例中，基于所述张力张量矩阵、所述涡度值计算并确定所述目标数据点的涡旋强度；可以理解的是，由于已公开相应的涡旋强度公式，那么涡旋强度Q值的公式中，由于，因此，，进而计算并确定目标数据点的涡旋强度。其中，表示涡度向量的模。In this embodiment, the vortex strength of the target data point is calculated and determined based on the tension tensor matrix and the vorticity value; it can be understood that since the corresponding vortex strength formula has been published, in the formula of the vortex strength Q value, ,therefore, , and then calculate and determine the vortex intensity of the target data point. Among them, Represents the magnitude of the vorticity vector.
本实施例中，将所述涡旋强度、所述涡度值、所述雅各比矩阵的F范数、所述张力张量矩阵的F范数作为各所述涡区域的物理属性信息。可以理解的是，通过上述各个公式分别计算得到相应的属性信息，将涡旋强度、涡度值、雅各比矩阵的F范数、张力张量矩阵的F范数作为各涡区域的物理属性信息，这样一来，通过各物理属性信息从不同角度反应了涡的特征，全面地刻画涡区域的性质，避免仅依赖单一指标可能导致的片面性，而且多种物理属性的结合可以提供更丰富的信息，有助于后续更准确地确定涡区域事件类型，减少误判的可能性。In this embodiment, the vortex intensity, the vorticity value, the F norm of the Jacobi matrix, and the F norm of the tension tensor matrix are used as the physical property information of each vortex region. It can be understood that the corresponding property information is calculated by the above formulas respectively, and the vortex intensity, vorticity value, the F norm of the Jacobi matrix, and the F norm of the tension tensor matrix are used as the physical property information of each vortex region. In this way, the characteristics of the vortex are reflected from different angles through each physical property information, and the properties of the vortex region are fully characterized, avoiding the one-sidedness that may be caused by relying on a single indicator. Moreover, the combination of multiple physical properties can provide richer information, which is helpful to more accurately determine the event type of the vortex region in the future and reduce the possibility of misjudgment.
步骤S14：基于各所述涡区域的物理属性信息分别构建所述涡区域的属性矩阵，并利用预设主成分分析法对各所述属性矩阵进行矩阵提取，以分别得到各所述涡区域的目标属性矩阵。Step S14: constructing attribute matrices of the vortex regions respectively based on the physical attribute information of the vortex regions, and performing matrix extraction on the attribute matrices using a preset principal component analysis method to obtain target attribute matrices of the vortex regions respectively.
本实施例中，按照所述目标数据点的所述涡旋强度、所述涡度值、所述雅各比矩阵的F范数、所述张力张量矩阵的F范数、三维空间坐标信息的顺序构建所述目标数据点的属性向量；基于位于同一所述涡区域的目标数据点的属性向量构建的特征属性矩阵作为所述涡区域的属性矩阵，以得到各所述涡区域的属性矩阵。可以理解的是，将每个涡区域的各数据点的涡旋强度Q、涡度值W，雅各比矩阵的F范数，张力张量矩阵的F范数，三维坐标信息按顺序构成一个描述该数据点的属性向量（Q，W，，，x，y，z）。然后将同一涡区域的目标数据点的属性向量作为矩阵元素，构建涡区域的属性矩阵，例如：涡区域中有m个数据点，那么根据m个数据点的属性向量构成m×7的涡区域的属性矩阵X。In this embodiment, the attribute vector of the target data point is constructed in the order of the vortex intensity, the vorticity value, the F norm of the Jacobian matrix, the F norm of the tension tensor matrix, and the three-dimensional space coordinate information of the target data point; the characteristic attribute matrix constructed based on the attribute vector of the target data point located in the same vortex region is used as the attribute matrix of the vortex region to obtain the attribute matrix of each vortex region. It can be understood that the vortex intensity Q, vorticity value W, F norm of the Jacobian matrix of each data point in each vortex region are , the F-norm of the tension tensor matrix , three-dimensional coordinate information Construct an attribute vector (Q, W, , , x, y, z). Then the attribute vectors of the target data points in the same vortex region are used as matrix elements to construct the attribute matrix of the vortex region. For example, if there are m data points in the vortex region, then the attribute matrix X of the vortex region of m×7 is constructed according to the attribute vectors of the m data points.
本实施例中，对所述属性矩阵中的属性元素进行标准化处理，以得到标准化属性矩阵。可以理解的是，对属性矩阵每列进行标准化，标准化的目的是使得每个特征的均值为0和标准差为1，避免不同物理属性信息之间的数据差异影响结果。对于属性矩阵X中的每个物理属性，计算标准化后的值：In this embodiment, the attribute elements in the attribute matrix are standardized to obtain a standardized attribute matrix. It can be understood that the purpose of standardizing each column of the attribute matrix is to make the mean of each feature 0 and the standard deviation 1, so as to avoid the data differences between different physical attribute information affecting the results. , calculate the standardized value:
； ;
其中，表示矩阵中所有列的元素均值，表示矩阵中所有列的元素标准差。对均值标准差标准化处理完成后得到的矩阵作为标准化属性矩阵。in, represents the mean of all columns in the matrix. Represents the standard deviation of all columns in the matrix. The matrix obtained after standardizing the mean and standard deviation is used as the standardized attribute matrix.
本实施例中，根据所述标准化属性矩阵计算对应的协方差矩阵；In this embodiment, the corresponding covariance matrix is calculated according to the standardized attribute matrix;
； ;
其中，当前的X为标准化属性矩阵，XT为标准化属性矩阵的转置，n表示矩阵行数，矩阵乘法XTX的结果是一个7×7的矩阵，每个矩阵元素Cjk表示特征j和特征k之间的协方差，计算如下：Among them, the current X is the standardized attribute matrix,XT is the transpose of the standardized attribute matrix, n represents the number of matrix rows, and the result of matrix multiplicationXTX is a 7×7 matrix. Each matrix elementCjk represents the covariance between feature j and feature k, which is calculated as follows:
； ;
其中，表示特征j的物理属性，表示特征k的物理属性。in, represents the physical properties of feature j, Represents the physical properties of feature k.
通过上述公式逐个计算每个矩阵元素的协方差值，以得到协方差矩阵。The covariance value of each matrix element is calculated one by one using the above formula to obtain the covariance matrix.
本实施例中，利用协方差矩阵构建对应的投影矩阵，以根据所述投影矩阵和所述标准化属性矩阵确定各所述涡区域的目标属性矩阵。具体的，对所述协方差矩阵进行特征值分解，以计算所述协方差矩阵的特征值和相应的特征向量，并从所述特征值中筛选目标特征值和相应的目标特征向量构建投影矩阵。可以理解的是，对协方差矩阵C进行特征值分解，计算协方差矩阵的特征值和相应的特征向量。特征值反映了每个特征向量方向上的方差量。每个特征值都对应一个特征向量。将特征值按照从大到小的顺序排序，选择前4个最大的特征值及其对应的特征向量作为主成分组成投影矩阵P，P的大小为7×4。In this embodiment, the covariance matrix is used to construct the corresponding projection matrix to determine the target attribute matrix of each vortex area according to the projection matrix and the standardized attribute matrix. Specifically, the covariance matrix is subjected to eigenvalue decomposition to calculate the eigenvalues and corresponding eigenvectors of the covariance matrix, and the target eigenvalues and corresponding target eigenvectors are selected from the eigenvalues to construct the projection matrix. It can be understood that the covariance matrix C is subjected to eigenvalue decomposition to calculate the eigenvalues and corresponding eigenvectors of the covariance matrix. The eigenvalue reflects the amount of variance in the direction of each eigenvector. Each eigenvalue corresponds to an eigenvector. The eigenvalues are sorted in order from large to small, and the first 4 largest eigenvalues and their corresponding eigenvectors are selected as the principal components to form the projection matrix P, and the size of P is 7×4.
本实施例中，利用所述投影矩阵对所述标准化属性矩阵进行投影降维，以得到相应的主成分属性矩阵，并将各所述主成分属性矩阵作为各所述涡区域的目标属性矩阵。可以理解的是，使用投影矩阵P将原始数据标准化后的X投影到新的低维空间中。具体投影公式如下所示：In this embodiment, the projection matrix is used to project the standardized attribute matrix to reduce the dimension, so as to obtain the corresponding principal component attribute matrix, and each principal component attribute matrix is used as the target attribute matrix of each vortex region. It can be understood that the projection matrix P is used to project the X after the original data is standardized into a new low-dimensional space. The specific projection formula is as follows:
； ;
其中，Xreduced是一个大小为m×4的矩阵，Ti时间步有n个涡特征区域，即有n个m×4的主成分属性矩阵，也即目标属性矩阵。Among them, Xreduced is a matrix of size m×4, and there are n vortex feature areas in theTi time step, that is, there are n m×4 principal component attribute matrices, which are also the target attribute matrices.
步骤S15：根据相邻时间步下的所述目标属性矩阵之间的相似度计算结果构建特征相似度矩阵，根据预设特征对应关系遍历所述特征相似度矩阵，以判断所述三维时序流场在不同时间步之间变化中所属的目标流场事件类型。Step S15: construct a feature similarity matrix according to the similarity calculation results between the target attribute matrices at adjacent time steps, and traverse the feature similarity matrix according to the preset feature correspondence relationship to determine the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps.
本实施例中，根据相邻时间步下的目标属性矩阵之间的相似度计算结果，然后通过相似度计算结果构建特征相似度矩阵，其中，特征相似度矩阵表征了两个相邻涡区域之间是否相似，然后通过遍历特征相似度矩阵，以及预先设置的特征对应关系，确定遍历后的特征数量属于哪一特征对应关系，进而确定出三维时序流场在不同时间步变化过程中，所属的目标流场事件类型。In this embodiment, based on the similarity calculation results between the target attribute matrices at adjacent time steps, a feature similarity matrix is constructed using the similarity calculation results, wherein the feature similarity matrix represents whether two adjacent vortex regions are similar. Then, by traversing the feature similarity matrix and the pre-set feature correspondence relationship, it is determined to which feature correspondence the traversed feature quantity belongs, thereby determining the target flow field event type to which the three-dimensional time series flow field belongs during the change process at different time steps.
可见，本申请公开了获取三维时序流场的各时间步下的流场数据；识别各时间步下的所述流场数据中属于涡区域的数据点，并基于所有所述数据点对各时间步下的流场区域进行空间联通区域分割，以得到单时间步下的涡区域；计算各所述涡区域的各数据点对应的涡旋强度、涡度值、局部剪切率，以得到各所述涡区域的物理属性信息；基于各所述涡区域的物理属性信息分别构建所述涡区域的属性矩阵，并利用预设主成分分析法对各所述属性矩阵进行矩阵提取，以分别得到各所述涡区域的目标属性矩阵；根据相邻时间步下的所述目标属性矩阵之间的相似度计算结果构建特征相似度矩阵，根据预设特征对应关系遍历所述特征相似度矩阵，以判断所述三维时序流场在不同时间步之间变化中所属的目标流场事件类型。由此可见，通过对于各时间步下的流场区域进行区域分割，进而获取到单个时间步下的涡区域，以便根据涡区域中的流场数据计算其物理属性信息，然后对物理属性信息进行融合简化，获取目标属性矩阵，最后通过相邻时间步下的目标属性矩阵的相似度结果进一步构建特征相似度矩阵，对特征相似度矩阵与预设特征对应关系进行匹配，实现三维流场涡特征间的匹配和事件跟踪，减少了阈值依赖，提高了计算效率。It can be seen that the present application discloses the acquisition of flow field data at each time step of a three-dimensional time series flow field; identification of data points belonging to the vortex region in the flow field data at each time step, and spatially connected region segmentation of the flow field region at each time step based on all the data points to obtain the vortex region at a single time step; calculation of the vortex intensity, vorticity value, and local shear rate corresponding to each data point in each vortex region to obtain physical property information of each vortex region; construction of attribute matrices of the vortex region based on the physical property information of each vortex region, and matrix extraction of each attribute matrix using a preset principal component analysis method to obtain target attribute matrices of each vortex region; construction of a feature similarity matrix based on the similarity calculation results between the target attribute matrices at adjacent time steps, and traversal of the feature similarity matrix according to the preset feature correspondence to determine the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps. It can be seen that by segmenting the flow field area at each time step, the vortex area at a single time step is obtained, so that the physical property information can be calculated according to the flow field data in the vortex area, and then the physical property information is fused and simplified to obtain the target attribute matrix. Finally, the feature similarity matrix is further constructed through the similarity results of the target attribute matrix at adjacent time steps, and the feature similarity matrix is matched with the preset feature correspondence to achieve matching and event tracking between three-dimensional flow field vortex features, reduce threshold dependence, and improve computational efficiency.
参照图5所示，本发明实施例公开了一种具体的基于物理信息融合的流场特征提取跟踪方法，相对于上一实施例，本实施例对技术方案作了进一步的说明和优化。具体的：As shown in FIG5 , the embodiment of the present invention discloses a specific flow field feature extraction and tracking method based on physical information fusion. Compared with the previous embodiment, this embodiment further illustrates and optimizes the technical solution. Specifically:
步骤S21：获取三维时序流场的各时间步下的流场数据。Step S21: Obtain flow field data at each time step of the three-dimensional time series flow field.
步骤S22：识别各时间步下的所述流场数据中属于涡区域的数据点，并基于所有所述数据点对各时间步下的流场区域进行空间联通区域分割，以得到单时间步下的涡区域。Step S22: Identify data points belonging to the vortex region in the flow field data at each time step, and perform spatially connected region segmentation on the flow field region at each time step based on all the data points to obtain the vortex region at a single time step.
步骤S23：计算各所述涡区域的各数据点对应的涡旋强度、涡度值、局部剪切率，以得到各所述涡区域的物理属性信息。Step S23: Calculate the vortex intensity, vorticity value, and local shear rate corresponding to each data point of each vortex region to obtain physical property information of each vortex region.
步骤S24：基于各所述涡区域的物理属性信息分别构建所述涡区域的属性矩阵，并利用预设主成分分析法对各所述属性矩阵进行矩阵提取，以分别得到各所述涡区域的目标属性矩阵。Step S24: constructing attribute matrices of the vortex regions respectively based on the physical attribute information of each vortex region, and performing matrix extraction on each attribute matrix using a preset principal component analysis method to obtain target attribute matrices of each vortex region respectively.
其中，步骤S21至步骤S24中更加详细的处理过程请参照前述公开的实施例内容，在此不再进行赘述。For more detailed processing procedures in step S21 to step S24, please refer to the aforementioned disclosed embodiment contents, which will not be described again here.
步骤S25：计算相邻时间步下的所述目标属性矩阵之间的余弦相似度；对所述余弦相似度取平均值，以得到相邻时间步下的所述目标属性矩阵之间的相似度计算结果。Step S25: Calculate the cosine similarities between the target attribute matrices at adjacent time steps; average the cosine similarities to obtain similarity calculation results between the target attribute matrices at adjacent time steps.
本实施例中，对Ti、Ti+1相邻两个时间步的所有目标属性矩阵求两两之间相似度。具体步骤如下：In this embodiment, the similarity between all target attribute matrices in two adjacent time stepsTi and Ti+1 is calculated. The specific steps are as follows:
计算Ti时间步中ni个目标属性矩阵Xi(i=1,2,...,ni)与Ti+1时间步中ni+1个目标属性矩阵Yj(j=1,2,...,n_(i+1))的相似度。Calculate the similarity between the ni target attribute matricesXi (i=1,2,...,ni ) in theTi time step and the ni+1 target attribute matrices Yj (j=1,2,...,n_(i+1)) in the Ti+1 time step.
； ;
其中，和分别是矩阵Xi和Yj中的第l和k行。in, and are the lth and kth rows in matricesXi andYj respectively.
计算平均余弦相似度，通过计算这些余弦值的平均值来得到两个矩阵之间的平均余弦相似度：Calculate the average cosine similarity between the two matrices by averaging these cosine values:
； ;
其中，矩阵Xi和Yj分别有m和p行。Here, matricesXi andYj have m and p rows respectively.
步骤S26：判断所述相似度计算结果与预设最小相似阈值之间的大小关系，若所述相似度计算结果小于所述预设最小相似阈值，则表征当前相邻时间步的特征区域不相似，并将当前相似度计算结果赋值0，以得到相应的第一赋值结果，然后将所述第一赋值结果作为构建特征相似度矩阵的矩阵元素；若所述相似度计算结果大于或等于所述预设最小相似阈值，则表征当前相邻时间步的特征区域相似，并将当前相似度计算结果赋值1，以得到相应的第二赋值结果，然后将所述第二赋值结果作为构建特征相似度矩阵的矩阵元素；利用所有所述相似度计算结果赋值后的赋值结果构建所述特征相似度矩阵。Step S26: Determine the size relationship between the similarity calculation result and the preset minimum similarity threshold. If the similarity calculation result is less than the preset minimum similarity threshold, it indicates that the feature regions of the current adjacent time steps are not similar, and the current similarity calculation result is assigned 0 to obtain a corresponding first assignment result, and then the first assignment result is used as a matrix element for constructing a feature similarity matrix; if the similarity calculation result is greater than or equal to the preset minimum similarity threshold, it indicates that the feature regions of the current adjacent time steps are similar, and the current similarity calculation result is assigned 1 to obtain a corresponding second assignment result, and then the second assignment result is used as a matrix element for constructing a feature similarity matrix; the feature similarity matrix is constructed using the assignment results after all the similarity calculation results are assigned.
本实施例中，根据相似度构建一个的特征相似矩阵，矩阵的行号表示Ti时间步的特征区域的编号，矩阵的列号表示Ti+1时间步的特征区域的编号。矩阵元素为0或1，0表示两个特征区域相似度小于最小相似阈值，即两个特征区域不相似。1表示两个特征区域相似度大于最小相似阈值，即两个特征区域相似。如图2、图3所示特征进行计算会得到如图6所示的特征相似矩阵。In this embodiment, a The feature similarity matrix is a matrix with row numbers representing the number of feature regions at time stepTi , and column numbers representing the number of feature regions at time step Ti+1 . The matrix elements are 0 or 1, where 0 means that the similarity between two feature regions is less than the minimum similarity threshold, i.e., the two feature regions are not similar. 1 means that the similarity between two feature regions is greater than the minimum similarity threshold, i.e., the two feature regions are similar. Calculating the features shown in Figures 2 and 3 will result in a feature similarity matrix as shown in Figure 6.
步骤S27：构建不同流场事件类型与流场特征之间的对应关系，以得到包含多个目标特征对应关系的预设特征对应关系；其中，所述流场事件类型包括：延续事件类型、分裂事件类型、合并事件类型、产生事件类型、耗散事件类型。Step S27: constructing a correspondence between different flow field event types and flow field features to obtain a preset feature correspondence including multiple target feature correspondences; wherein the flow field event types include: continuation event type, split event type, merge event type, generation event type, and dissipation event type.
本实施例中，如图7所示，当流场特征数量对应关系为1对1时，构建对应的延续事件类型与流场特征数量的第一目标特征对应关系；当流场特征数量对应关系为1对多时，构建对应的分裂事件类型与流场特征数量的第二目标特征对应关系；当流场特征数量对应关系为多对1时，构建对应的合并事件类型与流场特征数量的第三目标特征对应关系；当流场特征数量对应关系为0对1时，构建对应的产生事件类型与流场特征数量的第四目标特征对应关系；当流场特征数量对应关系为1对0时，构建对应的耗散事件类型与流场特征数量的第五目标特征对应关系。根据第一目标特征对应关系、第二目标特征对应关系、第三目标特征对应关系、第四目标特征对应关系、第五目标特征对应关系构建预设特征对应关系。In this embodiment, as shown in FIG7 , when the correspondence between the number of flow field features is 1 to 1, a first target feature correspondence between the corresponding continuation event type and the number of flow field features is constructed; when the correspondence between the number of flow field features is 1 to many, a second target feature correspondence between the corresponding splitting event type and the number of flow field features is constructed; when the correspondence between the number of flow field features is many to 1, a third target feature correspondence between the corresponding merging event type and the number of flow field features is constructed; when the correspondence between the number of flow field features is 0 to 1, a fourth target feature correspondence between the corresponding generation event type and the number of flow field features is constructed; when the correspondence between the number of flow field features is 1 to 0, a fifth target feature correspondence between the corresponding dissipation event type and the number of flow field features is constructed. A preset feature correspondence is constructed according to the first target feature correspondence, the second target feature correspondence, the third target feature correspondence, the fourth target feature correspondence, and the fifth target feature correspondence.
步骤S28：遍历所述特征相似度矩阵，以根据矩阵遍历结果以及预设特征对应关系确定目标特征对应关系；基于所述目标特征对应关系判断三维时序流场在不同时间步之间变化中所属的目标流场事件类型。Step S28: traverse the feature similarity matrix to determine the target feature correspondence according to the matrix traversal result and the preset feature correspondence; based on the target feature correspondence, determine the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps.
本实施例中，遍历特征相似度矩阵的各个矩阵元素，然后根据遍历后的各矩阵元素与某一特征对应关系是对应关系，以判断三维时序流场在不同时间步之间变化中所属的目标流场事件类型。In this embodiment, each matrix element of the feature similarity matrix is traversed, and then the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps is determined based on the correspondence between each traversed matrix element and a certain feature.
参照图8所示，本发明公开了一种T时刻和T+1时刻的流场特征提取跟踪方法，步骤如下：首先获取T时刻和T+1时刻的流场数据，然后进入特征区域（涡区域）识别分割过程，具体的，通过分别计算T时刻和T+1时刻下的流场数据的Q值（涡旋强度）进而，根据Q值是否大于0来进行流场数据点是否属于涡区域的识别处理，需要注意的是，此时的Q值大于0的流场数据点组成的涡区域并非最后确定的涡区域，而是一个初始涡区域，然后采用三维区域增长算法对初始涡区域中的数据点进行相应处理，以得到最终的涡区域及其内部数据点，完成T时刻流场和T+1时刻流场的特征区域（涡区域）分割处理，以分别得到T时刻下的特征区域C1、C2、…、Cn以及T+1时刻的特征区域D1、D2、…、Dm。然后对各特征区域进行物理属性计算，并根据计算出的物理属性信息构建属性矩阵，然后采用PCA主成分分析法将属性矩阵中的矩阵提取为相互独立的新物理属性，以去除物理属性之间的依赖，然后执行矩阵投影操作，以融合投影矩阵与标准化属性矩阵，得到主成分属性矩阵。其中，在融合属性矩阵的过程中，为了对T时刻和T+1时刻的融合过程加以区分，将T时刻主成分分析后得到的主成分属性矩阵记为X1,X2,…,Xn；将T+1时刻主成分分析后得到的主成分属性矩阵记为Y1,Y2,…,Ym。进一步的，对T时刻与T+1时刻的主成分属性矩阵之间进行余弦相似度的计算，并根据计算结果构建n×m特征相似度矩阵，最后通过预设特征对应关系并根据遍历特征相似度矩阵确定出T时刻到T+1时刻下之间变化中所属的目标流场事件类型，完成了流场事件的可视化跟踪。8 , the present invention discloses a flow field feature extraction and tracking method at time T and time T+1, and the steps are as follows: first, the flow field data at time T and time T+1 are obtained, and then a feature region (vortex region) identification and segmentation process is entered. Specifically, the Q value (vortex intensity) of the flow field data at time T and time T+1 is calculated respectively, and then, whether the flow field data point belongs to the vortex region is identified and processed according to whether the Q value is greater than 0. It should be noted that the vortex region composed of the flow field data points with a Q value greater than 0 at this time is not the final vortex region determined, but an initial vortex region. Then, a three-dimensional region growing algorithm is used to perform corresponding processing on the data points in the initial vortex region to obtain the final vortex region and its internal data points, and the feature region (vortex region) segmentation processing of the flow field at time T and the flow field at time T+1 is completed to obtain the feature regions C1, C2, …, Cn at time T and the feature regions D1, D2, …, Dm at time T+1 respectively. Then, the physical properties of each feature area are calculated, and the attribute matrix is constructed based on the calculated physical property information. Then, the PCA principal component analysis method is used to extract the matrix in the attribute matrix as independent new physical properties to remove the dependence between the physical properties. Then, the matrix projection operation is performed to fuse the projection matrix and the standardized attribute matrix to obtain the principal component attribute matrix. Among them, in the process of fusing the attribute matrix, in order to distinguish the fusion process at time T and time T+1, the principal component attribute matrix obtained after the principal component analysis at time T is recorded as X1, X2,…, Xn; the principal component attribute matrix obtained after the principal component analysis at time T+1 is recorded as Y1, Y2,…, Ym. Further, the cosine similarity between the principal component attribute matrices at time T and time T+1 is calculated, and an n×m feature similarity matrix is constructed based on the calculation results. Finally, the target flow field event type in the change between time T and time T+1 is determined by presetting the feature correspondence and traversing the feature similarity matrix, and the visual tracking of the flow field event is completed.
由此可见，通过求相邻时间步所有特征两两之间的余弦相似度，根据余弦相似度，大于最小阈值即认为两个特征相似，实现三维流场涡特征间的匹配，然后遍历两个矩阵中各元素两个特征是否相似的结果根据事件类型与流场特征之间的特征对应关系来判定所属的事件类型，然后按照时间顺序，逐个判断其所属的事件，实现了事件跟踪，减少了阈值依赖，提高了计算效率。It can be seen that by calculating the cosine similarity between all features in adjacent time steps, if the cosine similarity is greater than the minimum threshold, the two features are considered to be similar, and the matching between the three-dimensional flow field vortex features is realized. Then, the results of traversing each element in the two matrices to see if the two features are similar are used to determine the event type based on the feature correspondence between the event type and the flow field feature. Then, the events to which they belong are determined one by one in chronological order, thereby realizing event tracking, reducing threshold dependence, and improving computational efficiency.
参照图9所示，本发明还相应公开了一种基于物理信息融合的流场特征提取跟踪装置，包括：As shown in FIG. 9 , the present invention also discloses a flow field feature extraction and tracking device based on physical information fusion, including:
数据获取模块11，用于获取三维时序流场的各时间步下的流场数据；A data acquisition module 11 is used to acquire flow field data at each time step of a three-dimensional time series flow field;
区域分割模块12，用于识别各时间步下的所述流场数据中属于涡区域的数据点，并基于所有所述数据点对各时间步下的流场区域进行空间联通区域分割，以得到单时间步下的涡区域；A region segmentation module 12 is used to identify data points belonging to the vortex region in the flow field data at each time step, and to perform spatially connected region segmentation on the flow field region at each time step based on all the data points to obtain the vortex region at a single time step;
信息计算模块13，用于计算各所述涡区域的各数据点对应的涡旋强度、涡度值、局部剪切率，以得到各所述涡区域的物理属性信息；An information calculation module 13 is used to calculate the vortex intensity, vorticity value, and local shear rate corresponding to each data point of each vortex region to obtain physical property information of each vortex region;
矩阵构建模块14，用于基于各所述涡区域的物理属性信息分别构建所述涡区域的属性矩阵，并利用预设主成分分析法对各所述属性矩阵进行矩阵提取，以分别得到各所述涡区域的目标属性矩阵；A matrix construction module 14 is used to construct attribute matrices of the vortex regions based on the physical attribute information of each vortex region, and perform matrix extraction on each attribute matrix using a preset principal component analysis method to obtain a target attribute matrix of each vortex region;
跟踪模块15，用于根据相邻时间步下的所述目标属性矩阵之间的相似度计算结果构建特征相似度矩阵，根据预设特征对应关系遍历所述特征相似度矩阵，以判断所述三维时序流场在不同时间步之间变化中所属的目标流场事件类型。The tracking module 15 is used to construct a feature similarity matrix based on the similarity calculation results between the target attribute matrices at adjacent time steps, and traverse the feature similarity matrix according to the preset feature correspondence relationship to determine the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps.
可见，本申请公开了获取三维时序流场的各时间步下的流场数据；识别各时间步下的所述流场数据中属于涡区域的数据点，并基于所有所述数据点对各时间步下的流场区域进行空间联通区域分割，以得到单时间步下的涡区域；计算各所述涡区域的各数据点对应的涡旋强度、涡度值、局部剪切率，以得到各所述涡区域的物理属性信息；基于各所述涡区域的物理属性信息分别构建所述涡区域的属性矩阵，并利用预设主成分分析法对各所述属性矩阵进行矩阵提取，以分别得到各所述涡区域的目标属性矩阵；根据相邻时间步下的所述目标属性矩阵之间的相似度计算结果构建特征相似度矩阵，根据预设特征对应关系遍历所述特征相似度矩阵，以判断所述三维时序流场在不同时间步之间变化中所属的目标流场事件类型。由此可见，通过对于各时间步下的流场区域进行区域分割，进而获取到单个时间步下的涡区域，以便根据涡区域中的流场数据计算其物理属性信息，然后对物理属性信息进行融合简化，获取目标属性矩阵，最后通过相邻时间步下的目标属性矩阵的相似度结果进一步构建特征相似度矩阵，对特征相似度矩阵与预设特征对应关系进行匹配，实现三维流场涡特征间的匹配和事件跟踪，减少了阈值依赖，提高了计算效率。It can be seen that the present application discloses the acquisition of flow field data at each time step of a three-dimensional time series flow field; identification of data points belonging to the vortex region in the flow field data at each time step, and spatially connected region segmentation of the flow field region at each time step based on all the data points to obtain the vortex region at a single time step; calculation of the vortex intensity, vorticity value, and local shear rate corresponding to each data point in each vortex region to obtain physical property information of each vortex region; construction of attribute matrices of the vortex region based on the physical property information of each vortex region, and matrix extraction of each attribute matrix using a preset principal component analysis method to obtain target attribute matrices of each vortex region; construction of a feature similarity matrix based on the similarity calculation results between the target attribute matrices at adjacent time steps, and traversal of the feature similarity matrix according to the preset feature correspondence to determine the target flow field event type to which the three-dimensional time series flow field belongs in the change between different time steps. It can be seen that by segmenting the flow field area at each time step, the vortex area at a single time step is obtained, so that the physical property information can be calculated according to the flow field data in the vortex area, and then the physical property information is fused and simplified to obtain the target attribute matrix. Finally, the feature similarity matrix is further constructed through the similarity results of the target attribute matrix at adjacent time steps, and the feature similarity matrix is matched with the preset feature correspondence to achieve matching and event tracking between three-dimensional flow field vortex features, reduce threshold dependence, and improve computational efficiency.
进一步的，本申请实施例还公开了一种电子设备，图10是根据一示例性实施例示出的电子设备20结构图，图中的内容不能认为是对本申请的使用范围的任何限制。Furthermore, an embodiment of the present application also discloses an electronic device. FIG10 is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram cannot be regarded as any limitation on the scope of use of the present application.
图10为本申请实施例提供的一种电子设备20的结构示意图。该电子设备20，具体可以包括：至少一个处理器21、至少一个存储器22、电源23、通信接口24、输入输出接口25和通信总线26。其中，所述存储器22用于存储计算机程序，所述计算机程序由所述处理器21加载并执行，以实现前述任一实施例公开的基于物理信息融合的流场特征提取跟踪方法中的相关步骤。另外，本实施例中的电子设备20具体可以为电子计算机。FIG10 is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input/output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the flow field feature extraction and tracking method based on physical information fusion disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.
本实施例中，电源23用于为电子设备20上的各硬件设备提供工作电压；通信接口24能够为电子设备20创建与外界设备之间的数据传输通道，其所遵循的通信协议是能够适用于本申请技术方案的任意通信协议，在此不对其进行具体限定；输入输出接口25，用于获取外界输入数据或向外界输出数据，其具体的接口类型可以根据具体应用需要进行选取，在此不进行具体限定。In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.
其中，处理器21可以包括一个或多个处理核心，比如4核心处理器、8核心处理器等。处理器21可以采用DSP(Digital Signal Processing，数字信号处理)、FPGA(Field－Programmable Gate Array，现场可编程门阵列)、PLA(Programmable Logic Array，可编程逻辑阵列)中的至少一种硬件形式来实现。处理器21也可以包括主处理器和协处理器，主处理器是用于对在唤醒状态下的数据进行处理的处理器，也称CPU(Central ProcessingUnit，中央处理器)；协处理器是用于对在待机状态下的数据进行处理的低功耗处理器。在一些实施例中，处理器21可以在集成有GPU(Graphics Processing Unit，图像处理器)，GPU用于负责显示屏所需要显示的内容的渲染和绘制。一些实施例中，处理器21还可以包括AI(Artificial Intelligence，人工智能)处理器，该AI处理器用于处理有关机器学习的计算操作。Among them, the processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 21 can be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 21 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 21 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.
另外，存储器22作为资源存储的载体，可以是只读存储器、随机存储器、磁盘或者光盘等，其上所存储的资源可以包括操作系统221、计算机程序222等，存储方式可以是短暂存储或者永久存储。In addition, the memory 22, as a carrier for storing resources, can be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon can include an operating system 221, a computer program 222, etc., and the storage method can be temporary storage or permanent storage.
其中，操作系统221用于管理与控制电子设备20上的各硬件设备以及计算机程序222，以实现处理器21对存储器22中海量数据223的运算与处理，其可以是Windows Server、Netware、Unix、Linux等。计算机程序222除了包括能够用于完成前述任一实施例公开的由电子设备20执行的基于物理信息融合的流场特征提取跟踪方法的计算机程序之外，还可以进一步包括能够用于完成其他特定工作的计算机程序。数据223除了可以包括电子设备接收到的由外部设备传输进来的数据，也可以包括由自身输入输出接口25采集到的数据等。Among them, the operating system 221 is used to manage and control the hardware devices and computer programs 222 on the electronic device 20, so as to realize the operation and processing of the processor 21 on the massive data 223 in the memory 22, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the flow field feature extraction and tracking method based on physical information fusion performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can also further include a computer program that can be used to complete other specific tasks. In addition to data transmitted from an external device received by the electronic device, the data 223 can also include data collected by its own input and output interface 25, etc.
进一步的，本申请还公开了一种计算机可读存储介质，用于存储计算机程序；其中，所述计算机程序被处理器执行时实现前述公开的基于物理信息融合的流场特征提取跟踪方法。关于该方法的具体步骤可以参考前述实施例中公开的相应内容，在此不再进行赘述。Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the flow field feature extraction and tracking method based on physical information fusion disclosed above is implemented. For the specific steps of the method, reference can be made to the corresponding contents disclosed in the above embodiments, and no further description will be given here.
本说明书中各个实施例采用递进的方式描述，每个实施例重点说明的都是与其它实施例的不同之处，各个实施例之间相同或相似部分互相参见即可。对于实施例公开的装置而言，由于其与实施例公开的方法相对应，所以描述的比较简单，相关之处参见方法部分说明即可。In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
专业人员还可以进一步意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。结合本文中所公开的实施例描述的方法或算法的步骤可以直接用硬件、处理器执行的软件模块，或者二者的结合来实施。软件模块可以置于随机存储器RAM（Random Access Memory）、内存、只读存储器ROM（Read Only Memory）、电可编程EPROM（Electrically Programmable Read Only Memory）、电可擦除可编程EEPROM（ElectricErasable Programmable Read Only Memory）、寄存器、硬盘、可移动磁盘、CD-ROM（CompactDisc-Read Only Memory，紧凑型光盘只读储存器）、或技术领域内所公知的任意其它形式的存储介质中。Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application. The steps of the method or algorithm described in conjunction with the embodiments disclosed herein can be implemented directly with hardware, a software module executed by a processor, or a combination of the two. The software module can be placed in a random access memory RAM (Random Access Memory), memory, read-only memory ROM (Read Only Memory), electrically programmable EPROM (Electrically Programmable Read Only Memory), electrically erasable programmable EEPROM (ElectricErasable Programmable Read Only Memory), register, hard disk, removable disk, CD-ROM (CompactDisc-Read Only Memory), or any other form of storage medium known in the technical field.
最后，还需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者设备中还存在另外的相同要素。Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
以上对本发明所提供的方案进行了详细介绍，本文中应用了具体个例对本发明的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本发明的方法及其核心思想；同时，对于本领域的一般技术人员，依据本发明的思想，在具体实施方式及应用范围上均会有改变之处，综上所述，本说明书内容不应理解为对本发明的限制。The scheme provided by the present invention is introduced in detail above. Specific examples are used in this article to illustrate the principle and implementation mode of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation mode and application scope. In summary, the content of this specification should not be understood as limiting the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411140175.2ACN118657808B (en) | 2024-08-20 | 2024-08-20 | Flow field feature extraction and tracking method, device and equipment based on physical information fusion |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411140175.2ACN118657808B (en) | 2024-08-20 | 2024-08-20 | Flow field feature extraction and tracking method, device and equipment based on physical information fusion |
| Publication Number | Publication Date |
|---|---|
| CN118657808A CN118657808A (en) | 2024-09-17 |
| CN118657808Btrue CN118657808B (en) | 2024-10-18 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411140175.2AActiveCN118657808B (en) | 2024-08-20 | 2024-08-20 | Flow field feature extraction and tracking method, device and equipment based on physical information fusion |
| Country | Link |
|---|---|
| CN (1) | CN118657808B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119579646B (en)* | 2024-11-05 | 2025-07-25 | 重庆交通大学 | Vertical shaft vortex recognition and tracking algorithm |
| CN120337825B (en)* | 2025-06-19 | 2025-08-29 | 中国空气动力研究与发展中心计算空气动力研究所 | Aircraft aerodynamic simulation vortex characteristic analysis method, device, equipment and medium |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102867094A (en)* | 2012-09-19 | 2013-01-09 | 西安交通大学 | Establishment method for free surface flow model in moving particle semi-implicit algorithm |
| CN114722735A (en)* | 2022-03-15 | 2022-07-08 | 天津大学 | Flow field characteristic tracking method based on graph optimization |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6963810B2 (en)* | 2003-09-02 | 2005-11-08 | Tao Of Systems Integration, Inc. | Method and system for locating critical flow feature indicators in three dimensional flow regimes |
| CN117690501A (en)* | 2023-12-28 | 2024-03-12 | 中国人民解放军火箭军工程大学 | Vortex identification method and device |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102867094A (en)* | 2012-09-19 | 2013-01-09 | 西安交通大学 | Establishment method for free surface flow model in moving particle semi-implicit algorithm |
| CN114722735A (en)* | 2022-03-15 | 2022-07-08 | 天津大学 | Flow field characteristic tracking method based on graph optimization |
| Publication number | Publication date |
|---|---|
| CN118657808A (en) | 2024-09-17 |
| Publication | Publication Date | Title |
|---|---|---|
| CN118657808B (en) | Flow field feature extraction and tracking method, device and equipment based on physical information fusion | |
| Silver et al. | Volume tracking | |
| De Leeuw et al. | Collapsing flow topology using area metrics | |
| Post et al. | The state of the art in flow visualisation: Feature extraction and tracking | |
| US10083517B2 (en) | Segmentation of an image based on color and color differences | |
| Post et al. | Feature extraction and visualisation of flow fields. | |
| Takahashi et al. | Topological volume skeletonization and its application to transfer function design | |
| Obermaier et al. | Visual trends analysis in time-varying ensembles | |
| He et al. | Multivariate spatial data visualization: a survey | |
| Chandler et al. | Interpolation-based pathline tracing in particle-based flow visualization | |
| CN108520142B (en) | A method, device, equipment and storage medium for identifying the boundary of urban agglomeration | |
| Sane et al. | A survey of seed placement and streamline selection techniques | |
| CN118334363B (en) | Topographic feature semantic modeling method based on remote sensing image and LiDAR analysis | |
| Sanderson et al. | Analysis of recurrent patterns in toroidal magnetic fields | |
| CN119988465B (en) | Traffic network situation data display method and system | |
| CN119478278A (en) | A method, device, equipment and storage medium for parallel explicit sampling of geological modeling | |
| CN119357302B (en) | Quick extraction method and device for longest confluence path of small drainage basin | |
| Vidanes et al. | Extending point-based deep learning approaches for better semantic segmentation in CAD | |
| de Leeuw et al. | Multi-level topology for flow visualization | |
| CN118551584A (en) | Scene automatic generation method and system based on digital twin | |
| WO2025030812A1 (en) | Chip layout and training method, electronic device, storage medium, and chip system | |
| WO2016132318A1 (en) | Method and apparatus for modeling, visualization and analysis of materials | |
| CN111598239B (en) | Method and device for extracting process system of article based on graph neural network | |
| Fujishiro et al. | T-map: A topological approach to visual exploration of time-varying volume data | |
| US11586791B1 (en) | Visualization of data buses in circuit designs |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |