CN116110040A - Substation equipment identification method and device based on combination of SHOT features and ESF features - Google Patents
Substation equipment identification method and device based on combination of SHOT features and ESF featuresDownload PDFInfo
- Publication number
- CN116110040A CN116110040ACN202310176024.1ACN202310176024ACN116110040ACN 116110040 ACN116110040 ACN 116110040ACN 202310176024 ACN202310176024 ACN 202310176024ACN 116110040 ACN116110040 ACN 116110040A
- Authority
- CN
- China
- Prior art keywords
- scene
- point cloud
- model
- point
- key points
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/64—Three-dimensional objects
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/757—Matching configurations of points or features
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10028—Range image; Depth image; 3D point clouds
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
Description
Translated fromChinese技术领域technical field
本发明属于三维物体识别领域，具体设计了一种基于SHOT特征和ESF特征结合的变电站设备识别方法及装置。The invention belongs to the field of three-dimensional object identification, and specifically designs a substation equipment identification method and device based on the combination of SHOT features and ESF features.
背景技术Background technique
三维点云的物体识别工作需要从大量无序的三维点中找到属于要识别的物体点云集合。而三维点之间没有物理上的联系，要做到物体的识别，这需要借助特征的学习来完成识别的工作。The object recognition work of 3D point cloud needs to find the set of point cloud of the object to be recognized from a large number of unordered 3D points. However, there is no physical connection between three-dimensional points. To achieve object recognition, it is necessary to use feature learning to complete the recognition work.
但三维模型的特质使得构建一个逼真的三维模型是一项非常专业且复杂的工作，因此实现资源重用的三维模型检索技术成为了计算机视觉、计算机图形学领域的研究热点。However, the characteristics of 3D models make it a very professional and complicated task to construct a realistic 3D model. Therefore, 3D model retrieval technology to realize resource reuse has become a research hotspot in the fields of computer vision and computer graphics.
但是全局特征的识别需要对整个场景进行分割，但当面对复杂场景时，分割会难以进行，得不到想要识别的场景模型，从而难以进行物体识别。However, the recognition of global features requires segmentation of the entire scene, but when faced with complex scenes, segmentation will be difficult, and the scene model to be recognized cannot be obtained, making it difficult to perform object recognition.
发明内容Contents of the invention
本发明的目的在于克服现有技术中的不足，提供一种基于SHOT特征和ESF特征结合的变电站设备识别方法及装置，通过全局特征和局部特征对于物体识别的优缺点结合来实现对三维点云的识别，能够有效提高物体识别的精度。本发明提出的对于三维点云识别的方法不需要对场景进行分割，有效利用全局特征和局部特征的优点来提高识别的精度，有效提高变电站设备点云的识别精度。The purpose of the present invention is to overcome the deficiencies in the prior art, to provide a substation equipment identification method and device based on the combination of SHOT features and ESF features, and realize the three-dimensional point cloud by combining the advantages and disadvantages of global features and local features for object recognition The recognition can effectively improve the accuracy of object recognition. The method for three-dimensional point cloud recognition proposed by the present invention does not need to segment the scene, effectively utilizes the advantages of global features and local features to improve the recognition accuracy, and effectively improves the recognition accuracy of substation equipment point clouds.
为达到上述目的，本发明是采用下述技术方案实现的：In order to achieve the above object, the present invention is achieved by adopting the following technical solutions:
第一方面，本发明提供了一种基于SHOT特征和ESF特征结合的变电站设备识别方法，包括以下步骤：In the first aspect, the present invention provides a method for identifying substation equipment based on the combination of SHOT features and ESF features, comprising the following steps:
步骤a，获取变电站场景原始点云集合，进行去噪配准操作，得到变电站场景点云；Step a, obtain the original point cloud set of the substation scene, perform denoising and registration operation, and obtain the point cloud of the substation scene;
步骤b，根据变电站场景点云和模型库待识别的设备点云，进行点云法线和关键点的计算，得到场景点云和模型点云的关键点。Step b, according to the substation scene point cloud and the equipment point cloud to be recognized in the model library, calculate the point cloud normal and key points, and obtain the key points of the scene point cloud and model point cloud.
步骤c,根据得到的场景点云和模型点云的关键点，计算关键点处的SHOT特征，得到场景和模型的关键点的SHOT特征集合。Step c, according to the key points of the obtained scene point cloud and model point cloud, calculate the SHOT feature at the key point, and obtain the SHOT feature set of the key point of the scene and the model.
步骤d，根据得到的场景和模型的关键点处的SHOT特征，进行特征点的匹配，得到场景和模型匹配的关键点对。In step d, according to the obtained SHOT features at the key points of the scene and the model, the feature points are matched to obtain key point pairs for matching the scene and the model.
步骤e，根据得到的匹配的关键点对，进行Hough投票，得到模型在场景匹配到的实例点云集。In step e, Hough voting is performed according to the obtained matched key point pairs, and the instance point cloud set matched by the model in the scene is obtained.
步骤f，根据Hough投票结果进行是否下一步骤，如果没有匹配的实例存在，计算模型在场景的大致位置，得到模型在场景中的待判断点云集合。In step f, proceed to the next step according to the Hough voting result. If no matching instance exists, calculate the approximate position of the model in the scene, and obtain the point cloud set of the model in the scene to be judged.
步骤g，根据场景中的待判定点云集合和模型点云集合，进行ESF特征计算和比较相似度，得到模型在场景中匹配到的点云集，即变电站设备识别种类。In step g, according to the point cloud set to be determined and the model point cloud set in the scene, ESF feature calculation and similarity comparison are performed to obtain the point cloud set matched by the model in the scene, that is, the substation equipment identification type.
进一步的，步骤a中，获取变电站场景点云，包括：Further, in step a, the point cloud of the substation scene is obtained, including:
架设地面的三维激光扫描仪，对变电站进行扫描，分为多个站点，每个站点在变电站场景中架设标点，使用三维激光扫描仪自动扫描场景，通看仪器查看预扫描生成的点云场景，将整个变电站场景进行扫描；Set up a 3D laser scanner on the ground to scan the substation and divide it into multiple sites. Each site sets up punctuation points in the substation scene, uses the 3D laser scanner to automatically scan the scene, and checks the point cloud scene generated by the pre-scan through the instrument. Scan the entire substation scene;
将扫描数据导出配备专门软件进行处理，进行配准，去噪声，滤波处理，并对最终获取的点云数据输出成pcd格式，获得变电站场景点云。The scanned data is exported and equipped with special software for processing, registration, denoising, filtering, and the final obtained point cloud data is output into pcd format to obtain the point cloud of the substation scene.
进一步的，步骤b，进行点云法线和关键点的计算，得到场景点云和模型点云的关键点，包括：Further, in step b, calculate the point cloud normal and key points, and obtain the key points of the scene point cloud and model point cloud, including:
计算场景点云和模型点云点的法线，通过体素化网格实现下采样，从而计算出场景点云和模型点云的关键点。Calculate the normals of the scene point cloud and model point cloud points, and realize downsampling through the voxel grid, so as to calculate the key points of the scene point cloud and model point cloud.
进一步的，步骤c,计算该关键点处的SHOT特征，得到场景和模型的关键点的SHOT特征集合，包括：Further, step c, calculate the SHOT feature at the key point, and obtain the SHOT feature set of the key point of the scene and the model, including:
计算场景和模型点云关键点处的局部坐标系，通过奇异值分解得到xyz轴向量，确定xyz方向；Calculate the local coordinate system at the key points of the scene and model point cloud, obtain the xyz axis vector through singular value decomposition, and determine the xyz direction;
结合场景和模型关键点处的局部坐标系，计算场景和模型关键点处方位角、仰角、径向值，并每个分区分成11个间隔，最后归一化，得到场景和模型点云关键点处的352维SHOT特征。Combining the local coordinate system at the key points of the scene and the model, calculate the azimuth, elevation, and radial values of the key points of the scene and the model, and divide each partition into 11 intervals, and finally normalize to obtain the key points of the scene and model point cloud The 352-dimensional SHOT feature at .
进一步的，步骤d，进行特征点的匹配，得到场景和模型匹配的关键点对，包括：Further, in step d, match feature points to obtain key point pairs for scene and model matching, including:
使用对应分组方法寻找场景和设备关键点的匹配分组关系，通过KdTree搜索每个设备关键点在场景关键点中进行匹配，如果SHOT特征向量的欧式距离小于阈值，则认为两个是关键点是匹配的，以此来找到所有设备和场景的匹配的关键点对。其中遍历模型点云的所有关键点，对每个关键点通过KdTree搜索场景点云的关键点，进行SHOT特征的比较，通过欧式距离确定两个关键点的差值，看是否小于阈值，小于的话认为是一对的匹配点。Use the corresponding grouping method to find the matching grouping relationship between the key points of the scene and the device, and search for each key point of the device through KdTree to match among the key points of the scene. If the Euclidean distance of the SHOT feature vector is less than the threshold, it is considered that the two key points are matched. , so as to find matching keypoint pairs of all devices and scenes. It traverses all the key points of the model point cloud, searches the key points of the scene point cloud through KdTree for each key point, compares the SHOT features, and determines the difference between the two key points through the Euclidean distance to see if it is less than the threshold value. considered as a pair of matching points.
进一步的，步骤e，进行Hough投票，得到模型在场景匹配到的实例点云集，包括：Further, in step e, Hough voting is performed to obtain the instance point cloud set matched by the model in the scene, including:
Hough投票借助三维特征描述子，计算一系列的模型点和场景点的匹配对同时每一个模型特征点和模型形心都有一个相对位置关系，因此与此模型特征点相匹配的场景特征点能够对应出一个形心的位置，由此形心位置信息在霍夫空间内对相关参数进行投票。得到模型在场景存在的实例，最终得到模型在场景匹配到的实例点云集。Hough voting uses the three-dimensional feature descriptor to calculate a series of matching pairs of model points and scene points. At the same time, each model feature point has a relative position relationship with the model centroid, so the scene feature points that match this model feature point can be Corresponding to the position of a centroid, the centroid position information votes on the relevant parameters in the Hough space. Get the instances where the model exists in the scene, and finally get the instance point cloud set matched by the model in the scene.
进一步的，步骤f，根据Hough投票结果进行是否下一步骤，如果没有匹配的实例存在，计算模型在场景的大致位置，得到模型在场景中的待判断点云集合，包括：Further, in step f, whether to proceed to the next step according to the Hough voting result, if no matching instance exists, calculate the approximate position of the model in the scene, and obtain the point cloud set of the model in the scene to be judged, including:
假如SHOT特征不能识别在场景中识别到模型，则通过模型和场景点云关键点之间匹配关系，计算出场景中可能的模型点云集，即模型在场景中的待判断点云集合。If the SHOT feature cannot identify the model in the scene, the possible model point cloud set in the scene is calculated through the matching relationship between the model and the key points of the scene point cloud, that is, the point cloud set of the model in the scene to be judged.
进一步的，步骤g，进行ESF特征计算和比较相似度，得到模型在场景中匹配到的点云集，包括：Further, in step g, ESF feature calculation and similarity comparison are performed to obtain point cloud sets matched by the model in the scene, including:
确定设备在场景中可能存在的位置，计算其中点云物体的ESF特征描述子。对于场景中可能模型点云集和设备模型点云集，每次迭代，随机选择三个点，计算点距离，点距比，点面积和点角度4个函数，最终迭代大约2000次，得到ESF特征。ESF特征做对比，最终确定设备是否存在。Determine the possible location of the device in the scene, and calculate the ESF feature descriptor of the point cloud object in it. For the possible model point cloud set and equipment model point cloud set in the scene, three points are randomly selected for each iteration, and the four functions of point distance, point distance ratio, point area and point angle are calculated, and the final iteration is about 2000 times to obtain ESF features. ESF features are compared to determine whether the device exists.
第二方面，本发明提供了一种基于SHOT特征和ESF特征结合的变电站设备识别装置，包括：In the second aspect, the present invention provides a substation equipment identification device based on the combination of SHOT features and ESF features, including:
点云获取模块：用于获取变电站场景原始点云集合，进行去噪配准操作，得到变电站场景点云；Point cloud acquisition module: used to obtain the original point cloud collection of the substation scene, perform denoising and registration operations, and obtain the point cloud of the substation scene;
关键点模块：用于根据变电站场景点云和模型库待识别的设备点云，进行点云法线和关键点的计算，得到场景点云和模型点云的关键点；Key point module: used to calculate the point cloud normal and key points according to the substation scene point cloud and the equipment point cloud to be recognized in the model library, and obtain the key points of the scene point cloud and model point cloud;
特征计算模块：用于根据得到的场景点云和模型点云的关键点，计算关键点处的SHOT特征，得到场景和模型的关键点的SHOT特征集合；Feature calculation module: used to calculate the SHOT feature at the key point according to the key point of the obtained scene point cloud and model point cloud, and obtain the SHOT feature set of the key point of the scene and the model;
匹配模块：用于根据得到的场景和模型的关键点处的SHOT特征，进行特征点的匹配，得到场景和模型匹配的关键点对；Matching module: used to match the feature points according to the obtained scene and the SHOT feature at the key point of the model, and obtain the key point pair of scene and model matching;
Hough投票模块：用于根据得到的匹配的关键点对，进行Hough投票，得到模型在场景匹配到的实例点云集；Hough voting module: used to perform Hough voting according to the obtained matching key point pairs, and obtain the instance point cloud set matched by the model in the scene;
判断模块：用于根据Hough投票结果进行是否下一步骤，如果没有匹配的实例存在，计算模型在场景的大致位置，得到模型在场景中的待判断点云集合；Judgment module: used to perform the next step according to the Hough voting results. If no matching instance exists, calculate the approximate position of the model in the scene, and obtain the point cloud set of the model in the scene to be judged;
识别模块：用于根据场景中的待判定点云集合和模型点云集合，进行ESF特征计算和比较相似度，得到模型在场景中匹配到的点云集，即变电站设备识别种类。Recognition module: it is used to calculate ESF features and compare similarity according to the point cloud set to be determined and the model point cloud set in the scene, and obtain the point cloud set matched by the model in the scene, that is, the identification type of substation equipment.
第三方面，本发明提供一种基于SHOT特征和ESF特征结合的变电站设备识别装置，包括处理器及存储介质；In the third aspect, the present invention provides a substation equipment identification device based on the combination of SHOT features and ESF features, including a processor and a storage medium;
所述存储介质用于存储指令；The storage medium is used to store instructions;
所述处理器用于根据所述指令进行操作以执行根据第一方面所述方法的步骤。The processor is configured to operate in accordance with the instructions to perform the steps of the method according to the first aspect.
与现有技术相比，本发明所达到的有益效果：Compared with the prior art, the beneficial effects achieved by the present invention are as follows:
1、本发明通过全局特征和局部特征对于物体识别的结合来实现对三维点云的识别，能够有效提高物体识别的精度。1. The present invention realizes the recognition of three-dimensional point clouds through the combination of global features and local features for object recognition, which can effectively improve the accuracy of object recognition.
2、本发明提出的对于三维点云识别的方法不需要对场景进行分割，有效利用全局特征和局部特征的优点来提高识别的精度，有效提高变电站设备点云的识别精度。2. The method for 3D point cloud recognition proposed by the present invention does not need to segment the scene, effectively utilizes the advantages of global features and local features to improve the recognition accuracy, and effectively improves the recognition accuracy of substation equipment point clouds.
附图说明Description of drawings
图1是识别效果示意图(左侧为要识别的点云，右侧为场景点云，浅色框区域里为识别的设备)。Figure 1 is a schematic diagram of the recognition effect (the point cloud to be recognized is on the left, the scene point cloud is on the right, and the recognized device is in the light-colored frame area).
图2为本发明算法的流程图。Fig. 2 is a flowchart of the algorithm of the present invention.
具体实施方式Detailed ways
下面结合附图对本发明作进一步描述。以下实施例仅用于更加清楚地说明本发明的技术方案，而不能以此来限制本发明的保护范围。The present invention will be further described below in conjunction with the accompanying drawings. The following examples are only used to illustrate the technical solution of the present invention more clearly, but not to limit the protection scope of the present invention.
实施例一：Embodiment one:
本发明是采用通过SHOT局部特征描述子和ESF全局特征描述子结合的方式对变电站设备的三维点云进行识别。The present invention recognizes the three-dimensional point cloud of the substation equipment by combining the SHOT local feature descriptor and the ESF global feature descriptor.
三维点云的物体识别工作需要从大量无序的三维点中找到属于要识别的物体点云集合。而三维点之间没有物理上的联系，要做到物体的识别，这需要借助特征的学习来完成识别的工作。使用局部特征去完成三维点云物体的识别，首先需要计算点云特征点的局部特征描述子。然后进行特征之间的匹配和错误匹配点的去除。最后通过聚类或投票算法确定模型在场景中是否存在。通过二进制描述特征描述子可以减少特征描述子的数据大小，加快特征匹配的速度。但对于时间性能上的提升会带来识别精度的损失。也可以通过将点云颜色信息加入特征里面，增强对点云局部特征的描述来提高识别的精度。但当点云数据集的颜色信息并不丰富时，通过加入颜色特征并不会挺高识别的精度，反而会增加计算特征的时间。也可以使用全局特征进行三维点云的物体识别。使用全局特征进行识别首先需要对识别的场景进行分割，得到待识别的模型。然后计算全局特征进行模型和场景之间相似度的计算。视点特征直方图、聚类视点直方图使用直方图增强全局特征对于噪声的鲁棒性。形状函数集合通过不同形状函数描述点云物体，使特征更接近物体的几何特征。但是全局特征的识别需要对整个场景进行分割，但当面对复杂场景时，分割会难以进行，得不到想要识别的场景模型，从而难以进行物体识别。采用融合局部特征和全局特征的方法可以使点云识别避免因点云遮挡、点云缺失或点云结构重复等问题导致精度下降。The object recognition work of 3D point cloud needs to find the set of point cloud of the object to be recognized from a large number of unordered 3D points. However, there is no physical connection between three-dimensional points. To achieve object recognition, it is necessary to use feature learning to complete the recognition work. To use local features to complete the recognition of 3D point cloud objects, it is first necessary to calculate the local feature descriptors of point cloud feature points. Then the matching between features and the removal of wrong matching points are carried out. Finally, it is determined whether the model exists in the scene through clustering or voting algorithms. The data size of the feature descriptor can be reduced by binary description of the feature descriptor, and the speed of feature matching can be accelerated. However, the improvement of time performance will bring the loss of recognition accuracy. It is also possible to improve the recognition accuracy by adding the point cloud color information into the feature and enhancing the description of the local features of the point cloud. However, when the color information of the point cloud dataset is not rich, the accuracy of recognition will not be improved by adding color features, but it will increase the time to calculate features. It is also possible to use global features for object recognition in 3D point clouds. Recognition using global features first needs to segment the recognized scene to obtain the model to be recognized. Then calculate the global feature to calculate the similarity between the model and the scene. Viewpoint Feature Histogram, Clustering Viewpoint Histogram Use histograms to enhance the robustness of global features to noise. The set of shape functions describes the point cloud object through different shape functions, making the features closer to the geometric features of the object. However, the recognition of global features requires segmentation of the entire scene, but when faced with complex scenes, segmentation will be difficult, and the scene model to be recognized cannot be obtained, making it difficult to perform object recognition. Using the method of fusing local features and global features can make point cloud recognition avoid accuracy degradation due to problems such as point cloud occlusion, point cloud missing, or point cloud structure duplication.
图2描述了从点云输入，计算SHOT特征，对应分组计算匹配点、Hough投票，到计算ESF特征的过程，具体过程如下：Figure 2 describes the process from point cloud input, calculation of SHOT features, corresponding group calculation of matching points, Hough voting, and calculation of ESF features. The specific process is as follows:
架设地面的三维激光扫描仪，对变电站进行扫描，分为几个站点，每个站点在变电站场景中架设标点，使用三维激光扫描仪自动扫描场景，通看仪器查看预扫描生成的点云场景，将整个感兴趣的变电站场景进行扫描。Set up a 3D laser scanner on the ground to scan the substation and divide it into several stations. Each station sets up punctuation points in the substation scene, uses the 3D laser scanner to automatically scan the scene, and looks through the instrument to view the point cloud scene generated by pre-scanning. Scan the entire substation scene of interest.
将数据导出配备专门软件进行处理，进行配准，去噪声，滤波等处理。The data is exported and equipped with special software for processing, such as registration, denoising, filtering and other processing.
对最终获取的点云数据输出成pcd格式。Output the final acquired point cloud data into pcd format.
计算点云场景和模型的法线，计算三维点云中的法线是计算点云曲面点处的切平面的法向量。Calculate the normal of the point cloud scene and model, and calculate the normal in the 3D point cloud is to calculate the normal vector of the tangent plane at the surface point of the point cloud.
确定表面一点法线的问题近似于估计表面的一个相切问题，使用最小二乘法平面模拟估计，PCA主成分分析方法估计表面法线，通过分析协方差矩阵的特征矢量和特征值计算点云表面点的法线。根据公式Y＝UΣVT对协方差矩阵进行奇异值分解，取U中最后一列作为点云该点处的法向量。The problem of determining the normal of a point on the surface is similar to the problem of estimating the tangency of the surface, using the least squares method for plane simulation estimation, PCA principal component analysis method to estimate the surface normal, and calculating the point cloud surface by analyzing the eigenvectors and eigenvalues of the covariance matrix The point's normal. Singular value decomposition is performed on the covariance matrix according to the formula Y=UΣVT , and the last column in U is taken as the normal vector at this point of the point cloud.
对场景和要识别设备点云计算法线，要确定法线定向，对于点云中每个点p，计算p的指定半径R内的邻近点，计算在p点表面法线n，检查n的方向是否一致指向视点Vp，如果不是则反转。对所有法线
其中，
计算场景和要识别设备的关键点，使用体素化网格实现下采样的方法来得到关键点，将整个点云创建一个三维体素栅格，每个体素可以近似为一个三维立方体，立方体的大小根据点云的疏密度自己设置，在每个体素内，使用体素内所有点的重心来近似显示体素中的所有的点，这样既可以保留点云的表面特征，又可以减少点的数量。Calculate the key points of the scene and the device to be identified, use the voxel grid to realize the downsampling method to obtain the key points, and create a three-dimensional voxel grid from the entire point cloud, each voxel can be approximated as a three-dimensional cube, the cube The size is set according to the density of the point cloud. In each voxel, the center of gravity of all points in the voxel is used to approximate all the points in the voxel. This can not only preserve the surface characteristics of the point cloud, but also reduce the density of points. quantity.
首先将点云空间进行体素化，每个小的三维空间为一个体素，每个体素内包括点云内部的点的集合，对这些点加权平均得到重心替代体素内所有的点，体素内的求得的重心的这个点即为关键点。First, the point cloud space is voxelized. Each small three-dimensional space is a voxel. Each voxel contains a collection of points inside the point cloud. The weighted average of these points is used to obtain the center of gravity instead of all points in the voxel. The point of the center of gravity obtained in prime is the key point.
在得到关键点后，分别计算场景和设备在关键点处的SHOT特征，首先需要在关键点出构建局部坐标系，得到关键点处的指定半径R内邻近点。After the key points are obtained, the SHOT features of the scene and the device at the key points are calculated separately. First, a local coordinate system needs to be constructed at the key points to obtain the adjacent points within the specified radius R at the key points.
根据
其中，R为指定的半径，di为点pi到点p的距离。Among them, R is the specified radius, and di is the distance from point pi to point p.
并且考虑其邻近点到关键点的距离加入权重，di＝||pi-p||2。And consider the distance from the adjacent point to the key point and add the weight, di =||pi -p||2 .
其中，pi为邻近点i的坐标，p为关键点的坐标。Among them, pi is the coordinate of adjacent point i, and p is the coordinate of key point.
对M进行奇异值分解，得到三个特征向量，根据公式(1)(2)(3)确定x轴的正方向，同理可得z轴正方向，y轴由x轴和z轴叉乘得到。Singular value decomposition is performed on M to obtain three eigenvectors, and the positive direction of the x-axis is determined according to the formula (1)(2)(3). Similarly, the positive direction of the z-axis can be obtained, and the y-axis is cross-multiplied by the x-axis and the z-axis get.
其中，
计算每个邻近点和关键点的方位角、仰角、径向方向,得到关键点处的SHOT特征描述子的352特征直方图由8个方位角分区、2个高程分区和2个径向分区组成每个分区分成11个间隔，计算如下：Calculate the azimuth, elevation, and radial direction of each adjacent point and key point, and obtain the 352 feature histogram of the SHOT feature descriptor at the key point, which consists of 8 azimuth partitions, 2 elevation partitions and 2 radial partitions Each partition is divided into 11 intervals, calculated as follows:
对每一个子块构建特征直方图，计算子块内的点云表面的点的法向量np与其点处局部坐标系z轴方向的夹角余弦值为：cosθ＝np·z,在直方图的横轴[-1,1]区间上划分为11个间隔，根据局部表面上点的法向量与z轴的夹角余弦值，对相应的直方图进行统计，纵轴为点的个数。Construct a feature histogram for each sub-block, and calculate the cosine value of the angle between the normal vector np of the point cloud surface point in the sub-block and the z-axis direction of the local coordinate system at the point: cosθ=np z, in the histogram The horizontal axis [-1,1] of the graph is divided into 11 intervals, and the corresponding histogram is counted according to the cosine value of the angle between the normal vector of the point on the local surface and the z-axis, and the vertical axis is the number of points .
对每一个子块构建特征直方图，计算子块内的点云表面的点的法向量np与其点处局部坐标系z轴方向的夹角余弦值为：cosθ＝np·z,在直方图的横轴[-1,1]区间上划分为11个间隔，根据局部表面上点的法向量与z轴的夹角余弦值，对相应的直方图进行统计，纵轴为点的个数。Construct a feature histogram for each sub-block, and calculate the cosine value of the angle between the normal vector np of the point cloud surface point in the sub-block and the z-axis direction of the local coordinate system at the point: cosθ=np z, in the histogram The horizontal axis [-1,1] of the graph is divided into 11 intervals, and the corresponding histogram is counted according to the cosine value of the angle between the normal vector of the point on the local surface and the z-axis, and the vertical axis is the number of points .
为了解决构建直方图过程中的边界效应，使用四线性插值的方法解决。当每个点累加到直方图的特点间隔的时候，对直方图的相邻间隔以及子块直方图的对应间隔执行四线性插值。对于点与空间邻近的网格对应具有相同分区的直方图，对它进行四线性插值，每个计数乘以每个维度的权重为(1-d)，对于局部直方图来说，d为当前条目到分区中心值的距离，对于仰角和方位角来说，d是点到关键点的角距离，对于径向来说，d是点到关键点的欧式距离。沿着每个维度，d以直方图的单位来度量，即它由两个相邻的区间之间的距离进行归一化。In order to solve the boundary effect in the process of constructing the histogram, a four-linear interpolation method is used to solve it. As each point is accumulated to a characteristic interval of the histogram, quadlinear interpolation is performed on adjacent intervals of the histogram and corresponding intervals of the subblock histogram. For points and spatially adjacent grids corresponding to histograms with the same partition, it is quadlinearly interpolated, and each count is multiplied by the weight of each dimension (1-d), for local histograms, d is the current The distance from the entry to the center of the partition. For elevation and azimuth, d is the angular distance from the point to the keypoint. For radial, d is the Euclidean distance from the point to the keypoint. Along each dimension, d is measured in histogram units, i.e. it is normalized by the distance between two adjacent bins.
将整个直方图进行归一化，使其对点密度具有鲁棒性。Normalizes the entire histogram to be robust to point density.
计算场景点云和模型点云所有关键点的SHOT特征描述子。Calculate the SHOT feature descriptors of all key points in the scene point cloud and model point cloud.
使用对应分组方法寻找场景和设备关键点的匹配分组关系，通过KdTree搜索每个设备关键点在场景关键点中进行匹配，如果SHOT特征向量的欧式距离小于阈值，则认为两个是关键点是匹配的，以此来找到所有设备和场景的匹配的关键点对。其中遍历模型点云的所有关键点，对每个关键点通过KdTree搜索场景点云的关键点，进行SHOT特征的比较，通过欧式距离确定两个关键点的差值，看是否小于阈值，小于的话认为是一对的匹配点。Use the corresponding grouping method to find the matching grouping relationship between the key points of the scene and the device, and search for each key point of the device through KdTree to match among the key points of the scene. If the Euclidean distance of the SHOT feature vector is less than the threshold, it is considered that the two key points are matched. , so as to find matching keypoint pairs of all devices and scenes. It traverses all the key points of the model point cloud, searches the key points of the scene point cloud through KdTree for each key point, compares the SHOT features, and determines the difference between the two key points through the Euclidean distance to see if it is less than the threshold value. considered as a pair of matching points.
使用Hough投票寻找设备在场景中存在的依据，分为离线阶段和在线阶段：Use Hough voting to find the basis for the existence of equipment in the scene, which is divided into offline phase and online phase:
离线阶段：1.使用在SHOT特征描述子阶段计算得到的局部坐标系。Offline stage: 1. Use the local coordinate system calculated in the SHOT feature description sub-stage.
2.在设备模型中选出一个参考点，采用模型的重心作为参考点，设为CM。2. Select a reference point in the equipment model, use the center of gravity of the model as the reference point, and set it as CM .
3.计算设备中各个关键点
4.然后将上面基于全局坐标系计算出来的向量转换成局部坐标系向量
5.计算一个坐标系到另一个坐标系的变换5. Calculate the transformation from one coordinate system to another
在线阶段：1.对模型库中的关键点
2.设参考点CS与关键点
每个特征点与其相对于模型质心的相对位置相关联，使得每个对应的场景特征可以在三维Hough空间中进行投票，为当前场景中可能存在的质心位置积累证据。Each feature point is associated with its relative position with respect to the model centroid, so that each corresponding scene feature can be voted in the 3D Hough space, accumulating evidence for possible centroid positions in the current scene.
对于SHOT局部特征描述识别到的场景中的设备，通过ICP最近迭代点算法计算模型到场景中识别到的物体的旋转和平移矩阵，为后续工作的展开提供支持，ICP最近迭代点算法求解最优变换如下：For the equipment in the scene identified by the SHOT local feature description, the rotation and translation matrix from the model to the object identified in the scene is calculated by the ICP nearest iteration point algorithm to provide support for the development of follow-up work. The ICP nearest iteration point algorithm solves the optimal Transform as follows:
在场景和模型点云中，按照一定约束条件，寻找最邻近点，计算出最优匹配的参数R和t，使得误差函数最小，即为最优变换。误差函数如下：In the scene and model point cloud, according to certain constraints, find the nearest neighbor point, and calculate the optimal matching parameters R and t, so that the error function is minimized, which is the optimal transformation. The error function is as follows:
1.在模型点云P中取点集pi∈P1. Take the point set pi ∈ P in the model point cloud P
2.在场景点云Q中对应点集qi∈Q，使得||qi-pi||＝min2. In the scene point cloud Q, the corresponding point set qi ∈ Q, so that ||qi -pi ||=min
3.计算旋转矩阵R和平移矩阵t，使误差函数最小3. Calculate the rotation matrix R and translation matrix t to minimize the error function
4.对模型点云点集pi使用上一步求得的旋转矩阵R和平移矩阵t进行旋转和平移，得到新的对应点集p′i＝{Rpi+t,pi∈P}4. Use the rotation matrix R and translation matrixt obtained in the previous step to rotate and translate the point set p i of the model point cloud to obtain a new corresponding point set p′i ={Rpi +t,pi ∈ P}
5.计算p′i与对应点集qi的平均距离5. Calculate the average distance between p′i and the corresponding point set qi
6.如果d小于阈值或者大于预设的最大迭代次数，则停止迭代次数，否则返回第2步，直到满足收敛条件为止。6. If d is less than the threshold or greater than the preset maximum number of iterations, stop the number of iterations, otherwise return to step 2 until the convergence condition is met.
设置投票的阈值，对投票最多的实例通过icp配准的方法得到设备到场景中的旋转矩阵。如果实例数为0，则表明SHOT特征不能识别设备在场景是否存在。Set the voting threshold, and get the rotation matrix from the device to the scene through the icp registration method for the instance with the most votes. If the number of instances is 0, it indicates that the SHOT feature cannot identify whether the device exists in the scene.
计算设备点云的包围框右下点pmin和左上点pmax。Calculate the lower right point pmin and the upper left point pmax of the bounding box of the device point cloud.
计算匹配点对中设备模型的关键点和pmin，pmax的向量，根据旋转不变性的特性，计算得到场景中包围框的位置
计算描述点云全局特征的描述子，计算步骤如下：Calculate the descriptor describing the global features of the point cloud, the calculation steps are as follows:
每次迭代，随机选择3个点，计算点距离(D2)、点距比(D2ratio)、点面积(D3)和点角度(A3)四个函数，其中对于D2、D3和A3函数，每次迭代除了计算相应的距离、面积和角度外，还需要检测点对的连线是完全处于表面内(IN)、完全处于表面外(OUT)还是两者都有(MIXED)，最后根据检测结果，将计算的函数值分到3类直方图IN、OUT、MIXED其中的一个。For each iteration, 3 points are randomly selected to calculate the four functions of point distance (D2), point distance ratio (D2ratio), point area (D3) and point angle (A3), among which for D2, D3 and A3 functions, each time In addition to calculating the corresponding distance, area and angle, the iteration also needs to detect whether the connection of the point pair is completely inside the surface (IN), completely outside the surface (OUT) or both (MIXED). Finally, according to the detection result, Classify the calculated function values into one of the 3 types of histograms IN, OUT, MIXED.
通常，ESF迭代次数是2000次，最终得到分别表示三个角度、三个面积、三个距离和一个距离比形状函数的组合直方图，比较两个ESF特征直方图，来识别设备是否存在。Usually, the number of ESF iterations is 2000 times, and finally a combined histogram representing three angles, three areas, three distances and a distance ratio shape function is finally obtained, and two ESF feature histograms are compared to identify whether the device exists.
图1为识别效果示意图，图中左侧为要识别的点云，右侧为场景点云，浅色框区域里为识别的设备。Figure 1 is a schematic diagram of the recognition effect. The left side of the figure is the point cloud to be recognized, the right side is the scene point cloud, and the light-colored frame area is the recognized device.
本专利方法通过SHOT特征和ESF特征结合的方法去进行变电站设备的识别。SHOT特征是一种局部特征描述符，它使用奇异值分解来获得关键点处的局部坐标系，并将许多相邻点的方向作为正方向。通过这种方法获得的局部坐标系使得每次关键点的特征计算都相同，因此目标识别结果是唯一的。SHOT特征描述符利用径向、经度和维度对关键点的局部目标特征描述进行区域划分，可以有效地获取关键点周围的特征，对遮挡和杂波具有良好的识别能力。ESF特征是一个全局特征描述符，它使用距离、角度和面积函数来描述点云的整体特征。特征描述子具有很强的唯一性，对噪声和不完整表面具有鲁棒性。因此，使用SHOT特征和ESF特征去识别变电站设备点云堆点云遮挡、点云缺失、点云噪声等方面带来的识别问题具有优势。The patented method uses the method of combining SHOT features and ESF features to identify substation equipment. The SHOT feature is a local feature descriptor that uses singular value decomposition to obtain the local coordinate system at key points, and takes the orientation of many neighboring points as positive directions. The local coordinate system obtained by this method makes the feature calculation of each key point the same, so the target recognition result is unique. The SHOT feature descriptor uses radial, longitude, and latitude to divide the local target feature description of key points, which can effectively obtain features around key points, and has good recognition ability for occlusion and clutter. The ESF feature is a global feature descriptor that uses distance, angle, and area functions to describe the overall characteristics of point clouds. Feature descriptors are highly unique and robust to noise and incomplete surfaces. Therefore, it is advantageous to use SHOT features and ESF features to identify identification problems caused by substation equipment point cloud stack point cloud occlusion, point cloud missing, and point cloud noise.
实施例二：Embodiment two:
本实施例提供一种基于SHOT特征和ESF特征结合的变电站设备识别装置，包括：This embodiment provides a substation equipment identification device based on the combination of SHOT features and ESF features, including:
点云获取模块：用于获取变电站场景原始点云集合，进行去噪配准操作，得到变电站场景点云；Point cloud acquisition module: used to obtain the original point cloud collection of the substation scene, perform denoising and registration operations, and obtain the point cloud of the substation scene;
关键点模块：用于根据变电站场景点云和模型库待识别的设备点云，进行点云法线和关键点的计算，得到场景点云和模型点云的关键点；Key point module: used to calculate the point cloud normal and key points according to the substation scene point cloud and the equipment point cloud to be recognized in the model library, and obtain the key points of the scene point cloud and model point cloud;
特征计算模块：用于根据得到的场景点云和模型点云的关键点，计算关键点处的SHOT特征，得到场景和模型的关键点的SHOT特征集合；Feature calculation module: used to calculate the SHOT feature at the key point according to the key point of the obtained scene point cloud and model point cloud, and obtain the SHOT feature set of the key point of the scene and the model;
匹配模块：用于根据得到的场景和模型的关键点处的SHOT特征，进行特征点的匹配，得到场景和模型匹配的关键点对；Matching module: used to match the feature points according to the obtained scene and the SHOT feature at the key point of the model, and obtain the key point pair of scene and model matching;
Hough投票模块：用于根据得到的匹配的关键点对，进行Hough投票，得到模型在场景匹配到的实例点云集；Hough voting module: used to perform Hough voting according to the obtained matching key point pairs, and obtain the instance point cloud set matched by the model in the scene;
判断模块：用于根据Hough投票结果进行是否下一步骤，如果没有匹配的实例存在，计算模型在场景的大致位置，得到模型在场景中的待判断点云集合；Judgment module: used to perform the next step according to the Hough voting results. If no matching instance exists, calculate the approximate position of the model in the scene, and obtain the point cloud set of the model in the scene to be judged;
识别模块：用于根据场景中的待判定点云集合和模型点云集合，进行ESF特征计算和比较相似度，得到模型在场景中匹配到的点云集，即变电站设备识别种类。Recognition module: it is used to calculate ESF features and compare similarity according to the point cloud set to be determined and the model point cloud set in the scene, and obtain the point cloud set matched by the model in the scene, that is, the identification type of substation equipment.
本实施例的装置可以用于实现实施例一所述的方法。The device in this embodiment can be used to implement the method described in the first embodiment.
实施例三：Embodiment three:
本实施例提供一种基于SHOT特征和ESF特征结合的变电站设备识别装置，包括处理器及存储介质；This embodiment provides a substation equipment identification device based on the combination of SHOT features and ESF features, including a processor and a storage medium;
所述存储介质用于存储指令；The storage medium is used to store instructions;
所述处理器用于根据所述指令进行操作以执行根据实施例一所述方法的步骤。The processor is configured to operate according to the instructions to execute the steps of the method according to the first embodiment.
本领域内的技术人员应明白，本申请的实施例可提供为方法、系统、或计算机程序产品。因此，本申请可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
本申请是参照根据本申请实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。The present application is described with reference to flowcharts and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.
以上所述仅是本发明的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本发明技术原理的前提下，还可以做出若干改进和变形，这些改进和变形也应视为本发明的保护范围。The above is only a preferred embodiment of the present invention, it should be pointed out that for those of ordinary skill in the art, without departing from the technical principle of the present invention, some improvements and modifications can also be made. It should also be regarded as the protection scope of the present invention.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310176024.1ACN116110040A (en) | 2023-02-28 | 2023-02-28 | Substation equipment identification method and device based on combination of SHOT features and ESF features |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310176024.1ACN116110040A (en) | 2023-02-28 | 2023-02-28 | Substation equipment identification method and device based on combination of SHOT features and ESF features |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116110040Atrue CN116110040A (en) | 2023-05-12 |
Family
ID=86254265
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310176024.1APendingCN116110040A (en) | 2023-02-28 | 2023-02-28 | Substation equipment identification method and device based on combination of SHOT features and ESF features |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116110040A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116645517A (en)* | 2023-06-05 | 2023-08-25 | 广东电网有限责任公司 | Substation equipment identification method, device, equipment and medium |
| CN119049034A (en)* | 2024-08-07 | 2024-11-29 | 武汉理工大学 | Portal crane identification method and device, electronic equipment and storage medium |
| CN119313810A (en)* | 2024-09-12 | 2025-01-14 | 广东工业大学 | A method for constructing a three-dimensional model of a substation based on deep learning |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110686677A (en)* | 2019-10-10 | 2020-01-14 | 东北大学 | Global positioning method based on geometric information |
| CN111310845A (en)* | 2020-02-26 | 2020-06-19 | 广东电网有限责任公司电力科学研究院 | Substation equipment identification method, device and equipment |
| CN112669385A (en)* | 2020-12-31 | 2021-04-16 | 华南理工大学 | Industrial robot workpiece identification and pose estimation method based on three-dimensional point cloud characteristics |
- 2023
- 2023-02-28CNCN202310176024.1Apatent/CN116110040A/enactivePending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110686677A (en)* | 2019-10-10 | 2020-01-14 | 东北大学 | Global positioning method based on geometric information |
| CN111310845A (en)* | 2020-02-26 | 2020-06-19 | 广东电网有限责任公司电力科学研究院 | Substation equipment identification method, device and equipment |
| CN112669385A (en)* | 2020-12-31 | 2021-04-16 | 华南理工大学 | Industrial robot workpiece identification and pose estimation method based on three-dimensional point cloud characteristics |
Non-Patent Citations (1)
| Title |
|---|
| JUNFENG DING等: "Point cloud objective recognition method combining SHOT features and ESF features", 2022 INTERNATIONAL CONFERENCE ON CYBER-ENABLED DISTRIBUTED COMPUTING AND KNOWLEDGE DISCOVERY, 16 October 2022 (2022-10-16), pages 267 - 274, XP034325628, DOI: 10.1109/CyberC55534.2022.00052* |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116645517A (en)* | 2023-06-05 | 2023-08-25 | 广东电网有限责任公司 | Substation equipment identification method, device, equipment and medium |
| CN119049034A (en)* | 2024-08-07 | 2024-11-29 | 武汉理工大学 | Portal crane identification method and device, electronic equipment and storage medium |
| CN119049034B (en)* | 2024-08-07 | 2025-08-01 | 武汉理工大学 | Portal crane identification method and device, electronic equipment and storage medium |
| CN119313810A (en)* | 2024-09-12 | 2025-01-14 | 广东工业大学 | A method for constructing a three-dimensional model of a substation based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111553409B (en) | A Point Cloud Recognition Method Based on Voxel Shape Descriptors | |
| JP5705147B2 (en) | Representing 3D objects or objects using descriptors | |
| JP6216508B2 (en) | Method for recognition and pose determination of 3D objects in 3D scenes | |
| JP5677798B2 (en) | 3D object recognition and position and orientation determination method in 3D scene | |
| Zhu et al. | Single image 3d object detection and pose estimation for grasping | |
| CN116110040A (en) | Substation equipment identification method and device based on combination of SHOT features and ESF features | |
| Liu et al. | A novel rock-mass point cloud registration method based on feature line extraction and feature point matching | |
| Patterson IV et al. | Object detection from large-scale 3d datasets using bottom-up and top-down descriptors | |
| CN111145129A (en) | Point cloud denoising method based on hyper-voxels | |
| CN117148375A (en) | A fast loopback detection method for small viewing angle lidar in unstructured environment | |
| Shokoufandeh et al. | Many-to-many feature matching in object recognition: a review of three approaches | |
| Zong et al. | A fast and accurate planar-feature-based global scan registration method | |
| Arvanitis et al. | Broad-to-narrow registration and identification of 3D objects in partially scanned and cluttered point clouds | |
| Gong et al. | A two-level framework for place recognition with 3D LiDAR based on spatial relation graph | |
| You et al. | Canonical voting: Towards robust oriented bounding box detection in 3d scenes | |
| An et al. | Extracting statistical signatures of geometry and structure in 2D occupancy grid maps for global localization | |
| Jain et al. | 3D object recognition: Representation and matching | |
| CN119359774A (en) | A method for point cloud data registration | |
| Marton et al. | Part-Based Geometric Categorization and Object Reconstruction in Cluttered Table-Top Scenes: Paper Type: Categories (7) and (5) | |
| Tran et al. | 3D point cloud registration based on the vector field representation | |
| Lu et al. | Three-dimensional object recognition using an extensible local surface descriptor | |
| Li et al. | Automatic coarse registration of urban point clouds using line-planar semantic structural features | |
| CN117542008B (en) | Semantic point cloud fusion automatic driving scene identification method and storage medium | |
| CN118115737A (en) | A loop detection method and system based on semantic ellipsoid | |
| Cailliere et al. | 3D mirror symmetry detection using Hough transform |
Legal Events
| 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 |