技术领域technical field
本发明涉及数字图像处理领域，尤其是一种用于输电线路部件智能自动识别的方法。The invention relates to the field of digital image processing, in particular to a method for intelligent automatic identification of power transmission line components.
背景技术Background technique
输电线路是电网的骨架支撑，输电线路的正常运行与否，直接影响着整个国民经济的健康发展与人民的正常生活。通过无人机搭载专用巡检设备，对输电线路进行数据采集，利用图像处理技术分析采集数据，确定输电线路运行状态。一次无人机巡检任务可以获取输电线路大量的图像或视频信息，巡检结束后由巡检人员进行人工判读，确定输电线路部件位置及其存在的缺陷。由于数据量大，导致人工判读劳动强度大、耗时长，而且受巡检人员经验影响，经常出现部件与缺陷错判、漏判的情况。为了降低劳动强度，提高数据分析的准确率与时效性，亟需标准化、智能化的部件识别分析方法来实现巡检后输电线路巡检数据的自动处理。The transmission line is the skeleton support of the power grid. The normal operation of the transmission line directly affects the healthy development of the entire national economy and the normal life of the people. The UAV is equipped with special inspection equipment to collect data on the transmission line, and uses image processing technology to analyze and collect data to determine the operating status of the transmission line. A UAV inspection mission can obtain a large amount of image or video information of the transmission line. After the inspection is completed, the inspection personnel will conduct manual interpretation to determine the location of the transmission line components and their existing defects. Due to the large amount of data, manual interpretation is labor-intensive and time-consuming, and due to the experience of inspection personnel, misjudgments and missed judgments of parts and defects often occur. In order to reduce labor intensity and improve the accuracy and timeliness of data analysis, standardized and intelligent component identification and analysis methods are urgently needed to realize automatic processing of transmission line inspection data after inspection.
在输电线路部件自动识别方法中，大部分发明专利只注重某一部件的识别，如专利CN2013105246701.3、CN201510272155.5只是针对杆塔、鸟巢目标进行识别，没有对其余附属部件进行检测。专利CN201610906708.2利用Faster R-CNN方法对输电线路小部件进行了识别，但是输电线路附属部件在图像中占比差异较大，该方法没有考虑到图像的空间信息，最终输电线路部件识别的准确率有待改进与提高。In the automatic identification method of transmission line components, most invention patents only focus on the identification of a certain component, such as patents CN2013105246701.3 and CN201510272155.5, which only identify towers and bird’s nest targets, and do not detect other auxiliary components. Patent CN201610906708.2 uses the Faster R-CNN method to identify small parts of transmission lines, but the proportion of auxiliary parts of transmission lines in the image varies greatly. This method does not take into account the spatial information of the image, and the final recognition of transmission line components is accurate The rate needs to be improved and improved.
发明内容Contents of the invention
本发明的目的是提供一种用于输电线路部件智能自动识别的方法，有效的获取了巡检目标的空间信息，提高了输电线路目标识别的准确性能。The purpose of the present invention is to provide a method for intelligent automatic identification of transmission line components, which effectively obtains the spatial information of inspection targets and improves the accuracy of transmission line target identification.
为实现上述目的，本发明采用下述技术方案：To achieve the above object, the present invention adopts the following technical solutions:
一种用于输电线路部件智能自动识别的方法，包括以下步骤：A method for intelligent automatic identification of power transmission line components, comprising the following steps:
S1、巡检数据图像样本预处理步骤：以巡检数据原图像作为图像源，标记输电线路部件在原图像中的位置，并对每个输电线路部件添加属性标签，构建输电线路部件识别训练数据集合；S1. Preprocessing steps of inspection data image samples: use the original image of inspection data as the image source, mark the position of transmission line components in the original image, and add attribute labels to each transmission line component, and construct a transmission line component recognition training data set ;
S2、特征提取步骤：对完成预处理的巡检数据图像样本，利用卷积神经网络与特征金字塔网络，提取输电线路图像多层次特征；S2. Feature extraction step: for the preprocessed inspection data image sample, use convolutional neural network and feature pyramid network to extract multi-level features of transmission line image;
S3、训练目标定位回归网络与分类网络步骤：以提取到的输电线路图像多层次特征和标定的属性标签数据为训练输入数据，计算位置敏感得分图，并计算分类网络与目标定位回归网络的损失值，利用随机梯度下降法优化分类网络与回归网络的参数，从而实现训练数据中输电线路部件的最优分类与定位；S3. Steps of training target positioning regression network and classification network: use the extracted multi-level features of transmission line images and calibrated attribute label data as training input data, calculate the position-sensitive score map, and calculate the loss of classification network and target positioning regression network Value, using the stochastic gradient descent method to optimize the parameters of the classification network and regression network, so as to achieve the optimal classification and positioning of transmission line components in the training data;
S4、输电线路部件检测步骤：根据输电线路识别训练得到的训练参数，初始化检测网络，批量导入输电线路巡检数据，实现部件的自动定位与分类。S4. Transmission line component detection step: According to the training parameters obtained from the transmission line identification training, initialize the detection network, import transmission line inspection data in batches, and realize automatic positioning and classification of components.
进一步地，所述巡检数据图像样本预处理步骤引入数据扩容策略以扩展训练数据集合的容量，所述数据扩容策略包括镜像映射、平移、旋转、裁剪、尺度变换。Further, the preprocessing step of the inspection data image sample introduces a data expansion strategy to expand the capacity of the training data set, and the data expansion strategy includes mirror mapping, translation, rotation, cropping, and scale transformation.
进一步地，所述特征提取步骤具体包括以下步骤：Further, the feature extraction step specifically includes the following steps:
卷积网络处理步骤：将预处理的巡检数据图像样本作为训练输入数据集合，采用VGG-net16卷积网络中前13层网络，通过优化各层参数，实现输电线路图像特征的提取，得到512维的高层语意特征；Convolutional network processing steps: use the preprocessed inspection data image samples as the training input data set, use the first 13 layers of the VGG-net16 convolutional network, and optimize the parameters of each layer to realize the extraction of transmission line image features, and get 512 High-level semantic features of dimension;
特征金字塔网络处理步骤：特征金字塔网络处理步骤是在卷积网络中同时进行的，卷积网络是自下而上进行卷积计算，通过为每个网络阶段定义一个金字塔级别，选择每个网络阶段的最后一层输出作为特征图的参考集，然后自上而下对每个级别的参考集进行上采样，再把该特征横向与下一层参考集特征相连接。Feature pyramid network processing steps: The feature pyramid network processing steps are performed simultaneously in the convolutional network. The convolutional network performs convolution calculations from the bottom up. By defining a pyramid level for each network stage, each network stage is selected The output of the last layer is used as the reference set of the feature map, and then the reference set of each level is up-sampled from top to bottom, and then the feature is horizontally connected with the reference set feature of the next layer.
进一步地，所述预处理的巡检数据图像样本包括目标区域图像数据和类别标签数据。Further, the preprocessed inspection data image samples include target area image data and category label data.
进一步地，所述训练目标定位回归网络与分类网络步骤具体包括以下步骤：Further, the step of training the target positioning regression network and the classification network specifically includes the following steps:
构建位置敏感分数图步骤：在卷积网络与特征金字塔网络相结合提取特征图后，添加一个卷积得分层用于提取图像上每个类别的k2个位置敏感分数图，假设共有C+1个类别，经过卷积产生k2(C+1)个通道的输出层；Steps for constructing a position-sensitive score map: After the convolutional network and the feature pyramid network are combined to extract the feature map, a convolutional score layer is added to extract k2 position-sensitive score maps of each category on the image, assuming a total of C+ 1 category, after convolution to generate an output layer of k2 (C+1) channels;
构建目标定位回归网络步骤：采用固定比例的候选框提取方式对特征图进行候选区域提取，为实现对目标定位回归网络的训练，定义损失函数为：The step of constructing the target positioning regression network: using the fixed-ratio candidate frame extraction method to extract the candidate area of the feature map, in order to realize the training of the target positioning regression network, define the loss function as:
其中，i是区域提取框的序号，pi是区域提取框时目标概率，是真实标签的概率：若是正样本时为1，是负样本时为0，ti是预测边框的位置坐标，是真实边框的位置坐标，Lcls是分类损失函数，用来判别边框里是否含有目标；Lreg是回归函数，对边框做位置和大小的微调，Ncls是训练时mini-batch的大小，Nreg是区域提取框定位的数量，λ是平衡参数；Among them, i is the serial number of the region extraction frame, pi is the target probability when the region is extracted, is the probability of the real label: if it is a positive sample, it is 1, and if it is a negative sample, it is 0, ti is the position coordinate of the predicted frame, is the position coordinates of the real frame, Lcls is the classification loss function, used to judge whether there is a target in the frame; Lreg is a regression function, fine-tuning the position and size of the frame, Ncls is the size of the mini-batch during training, Nreg is the number of region extraction frame locations, and λ is a balance parameter;
为了实现对网络的整体训练，对卷积得分层、分类网络及目标定位回归网络进行参数随机初始化，定义损失函数为：In order to realize the overall training of the network, the parameters of the convolutional layer, classification network and target positioning regression network are randomly initialized, and the loss function is defined as:
其中，c*是候选区域的真实标签，[c*＞0]代表如果是真实标签时，参数设为1，否则为0；是用于分类的交叉熵损失函数，Lreg(t,t*)是边界回归损失函数，t*是真实标签的坐标，t是输出的回归坐标；Among them, c* is the real label of the candidate area, [c* > 0] means that if it is a real label, the parameter is set to 1, otherwise it is 0; is the cross-entropy loss function for classification, Lreg (t,t* ) is the boundary regression loss function, t* is the coordinate of the true label, and t is the regression coordinate of the output;
构建分类网络步骤：将感兴趣区域的标记框映射到卷积得分层上，并且划分k*k个分格，每个区域得到C+1个维度的特征图，按顺序依次将k*k个格子上对应的得分图抽取出来，然后这k2个得分图用来对候选区域进行投票从而判定目标的类别；Steps to build a classification network: map the marked frame of the region of interest to the convolutional score layer, and divide k*k grids, each region gets a feature map of C+1 dimensions, and sequentially map k*k The corresponding score maps on the grid are extracted, and then these k2 score maps are used to vote for the candidate area to determine the category of the target;
对于大小为w*h的感兴趣候选区域，每个分格的大小为(w/k)*(h/k)，对于第i行第j列的分格中的得分图，进行如下定义：For a candidate region of interest with a size of w*h, the size of each grid is (w/k)*(h/k), and for the score map in the grid of row i and column j, define as follows:
其中，rc(i,j)是针对第c类目标的第(i,j)个分格的响应得分，Zi,j,c代表k2(C+1)个得分图中的一个，(x0,y0)表示候选区域的左上角坐标，n表示每个格子中像素的个数，θ是网络中所有可学习的参数；Among them, rc (i, j) is the response score of the (i, j)th cell for the c-th category target, Zi, j, c represents one of the k2 (C+1) score maps, (x0 , y0 ) represents the coordinates of the upper left corner of the candidate area, n represents the number of pixels in each grid, and θ is all learnable parameters in the network;
对于k2个得分图通过平均得分进行投票，对于每个候选区域产生C+1维的向量：For k2 score maps, vote by the average score, and generate a C+1-dimensional vector for each candidate region:
最后计算每个目标的softmax响应：Finally compute the softmax response for each target:
进一步地，所述输电线路部件检测步骤包括以下步骤:Further, the transmission line component detection step includes the following steps:
初始化网络步骤：利用优化、训练得到的图像识别模型初始化训练网络，包括网络结构定义、网络基本参数定义、权重及偏置数据填充；Initialize the network step: use the optimized and trained image recognition model to initialize the training network, including network structure definition, network basic parameter definition, weight and bias data filling;
提取图像特征图步骤：首先，对输入图像进行多层卷积操作提取输入图像的高层语意表达；其次，对提取到的特征进行多尺度分割与上采样；最后，将金字塔特征与卷积后的特征进行融合，形成最终特征图；The steps of extracting image feature maps: first, perform multi-layer convolution operation on the input image to extract the high-level semantic expression of the input image; second, perform multi-scale segmentation and up-sampling on the extracted features; finally, combine the pyramid features with the convolved The features are fused to form the final feature map;
提取候选区域步骤：利用区域提取网络提取候选框，并进行非极大值抑制操作，只保留得分最高的前300个候选框，以实现检测的快速处理；The step of extracting candidate areas: use the area extraction network to extract candidate boxes, and perform non-maximum value suppression operations, and only keep the top 300 candidate boxes with the highest scores to achieve fast processing of detection;
计算位置敏感得分图步骤：对候选框进行得分图卷积操作，计算每个候选框的得分，并求取每个候选框的得分均值，构建位置敏感得分图；The step of calculating the position-sensitive score map: perform a score map convolution operation on the candidate frame, calculate the score of each candidate frame, and calculate the average score of each candidate frame, and construct a position-sensitive score map;
确定候选区域类别与标记框步骤：对敏感得分图进行softmax分类，确定最终类别；对候选框进行拟合优化确定最终目标标记位置。The step of determining the category of the candidate area and the marked frame: perform softmax classification on the sensitivity score map to determine the final category; perform fitting optimization on the candidate frame to determine the final target mark position.
发明内容中提供的效果仅仅是实施例的效果，而不是发明所有的全部效果，上述技术方案中的一个技术方案具有如下优点或有益效果：The effects provided in the summary of the invention are only the effects of the embodiments, rather than all the effects of the invention. One of the above technical solutions has the following advantages or beneficial effects:
本发明提供了一种自动化的巡检图像处理模式，能够减低巡检人员劳动强度、提高作业效率和智能化水平。通过金字塔特征提取网络，实现了巡检图像的多尺度特征提取，有效的获取了巡检目标的空间信息；利用位置敏感得分图，增强了检测模型对目标偏移的鲁棒性能。最终对改进的方法进行端对端的全局训练、调优，提高了输电线路目标识别的准确性能。The invention provides an automatic inspection image processing mode, which can reduce the labor intensity of inspection personnel and improve operation efficiency and intelligence level. Through the pyramid feature extraction network, the multi-scale feature extraction of the inspection image is realized, and the spatial information of the inspection target is effectively obtained; the position-sensitive score map is used to enhance the robustness of the detection model to the target deviation. Finally, end-to-end global training and tuning are performed on the improved method, which improves the accuracy of transmission line target recognition.
附图说明Description of drawings
图1是本发明用于输电线路部件智能自动识别的方法流程图；Fig. 1 is the flow chart of the method for intelligent automatic identification of power transmission line components according to the present invention;
图2是特征提取步骤流程图；Fig. 2 is a flow chart of feature extraction steps;
图3是结合不同层次特征金字塔特征网络原理图；Figure 3 is a schematic diagram of a feature network combining different levels of feature pyramids;
图4是基于位置敏感得分图的分类网络示意图；Figure 4 is a schematic diagram of a classification network based on a position-sensitive score map;
图5是训练目标定位回归网络与分类网络步骤流程图；Fig. 5 is the step flow chart of training target location regression network and classification network;
图6是输电线路部件检测步骤流程图。Fig. 6 is a flow chart of the detection steps of transmission line components.
具体实施方式detailed description
为了能清楚说明本方案的技术特点，下面通过具体实施方式，并结合其附图，对本发明进行详细阐述。为了简化本发明的公开，下文中对特定例子的部件和设置进行描述。此外，本发明可以在不同例子中重复参考数字和/或字母。这种重复是为了简化和清楚的目的，其本身不指示所讨论各种实施例和/或设置之间的关系。应当注意，在附图中所图示的部件不一定按比例绘制。本发明省略了对公知组件和处理技术及工艺的描述以避免不必要地限制本发明。In order to clearly illustrate the technical features of the present solution, the present invention will be described in detail below through specific implementation methods and in conjunction with the accompanying drawings. To simplify the disclosure of the present invention, components and arrangements of specific examples are described below. Furthermore, the present invention may repeat reference numerals and/or letters in different instances. This repetition is for the purpose of simplicity and clarity and does not in itself indicate a relationship between the various embodiments and/or arrangements discussed. It should be noted that components illustrated in the figures are not necessarily drawn to scale. Descriptions of well-known components and processing techniques and processes are omitted herein to avoid unnecessarily limiting the present invention.
如图1所示，一种用于输电线路部件智能自动识别的方法，包括以下步骤：As shown in Figure 1, a method for intelligent automatic identification of transmission line components includes the following steps:
S1、巡检数据图像样本预处理步骤：以无人机巡检数据原图像作为图像源，手动标记杆塔、绝缘子、均压环、间隔棒等输电线路部件在原图像中的位置，并对每个输电线路部件添加属性标签，构建输电线路部件识别训练数据集合；S1. Preprocessing steps of inspection data image samples: use the original image of UAV inspection data as the image source, manually mark the positions of transmission line components such as towers, insulators, voltage equalizing rings, spacers, etc. Add attribute labels to transmission line components to build a training data set for transmission line component recognition;
S2、特征提取步骤：对完成预处理的巡检数据图像样本，利用卷积神经网络与特征金字塔网络，提取输电线路图像多层次特征；S2. Feature extraction step: for the preprocessed inspection data image samples, use convolutional neural network and feature pyramid network to extract multi-level features of transmission line images;
S3、训练目标定位回归网络与分类网络步骤：以提取到的输电线路图像多层次特征和标定的属性标签数据为训练输入数据，计算位置敏感得分图，并计算分类网络与目标定位回归网络的损失值，利用随机梯度下降法优化分类网络与回归网络的参数，从而实现训练数据中输电线路部件的最优分类与定位；S3. Steps of training target positioning regression network and classification network: use the extracted multi-level features of transmission line images and calibrated attribute label data as training input data, calculate the position-sensitive score map, and calculate the loss of classification network and target positioning regression network Value, using the stochastic gradient descent method to optimize the parameters of the classification network and regression network, so as to achieve the optimal classification and positioning of transmission line components in the training data;
S4、输电线路部件检测步骤：根据输电线路识别训练得到的训练参数，初始化检测网络，批量导入输电线路巡检数据，实现部件的自动定位与分类。S4. Transmission line component detection step: According to the training parameters obtained from the transmission line identification training, initialize the detection network, import transmission line inspection data in batches, and realize automatic positioning and classification of components.
为了增加图像数据的多样性进而获取足够多的训练数据，执行步骤S1时引入数据扩容策略以扩展训练数据集合的容量，这里扩容策略设计的方法主要包括：镜像映射、平移、旋转、裁剪、尺度变换。五种数据扩容策略随机搭配使用，对手动标记的样本进行扩展，以满足训练数据量的要求。In order to increase the diversity of image data and obtain enough training data, a data expansion strategy is introduced to expand the capacity of the training data set when performing step S1. Here, the expansion strategy design methods mainly include: mirror mapping, translation, rotation, cropping, scaling transform. Five data expansion strategies are randomly used together to expand the manually marked samples to meet the requirements of the training data volume.
如图2所示，特征提取步骤具体包括以下步骤：As shown in Figure 2, the feature extraction step specifically includes the following steps:
S201、卷积网络处理步骤：将预处理的巡检数据图像样本作为训练输入数据集合，采用VGG-net16卷积网络中前13层网络，主要包括卷积层、池化层，通过优化各层参数，实现输电线路图像特征的提取，得到512维的高层语意特征。S201. Convolutional network processing step: use the preprocessed inspection data image samples as a training input data set, use the first 13 layers of the VGG-net16 convolutional network, mainly including convolutional layers and pooling layers, optimize each layer parameters to realize the extraction of transmission line image features and obtain 512-dimensional high-level semantic features.
S202、特征金字塔网络处理步骤：如图3所示，特征金字塔网络通过将低分辨率、高语义信息的高层特征与高分辨率、低语义信息的低层特征两种特征，进行自上而下的侧边连接，使得所有尺度下的特征都有丰富的语义信息，这种操作是在卷积网络中同时进行的，而不是采用构建图像金字塔的方式，从而可以避免额外的耗时和显存使用。卷积网络是自下而上进行卷积计算，通过为每个网络阶段定义一个金字塔级别，选择每个网络阶段的最后一层输出作为特征图的参考集，然后自上而下对每个级别的参考集进行上采样，再把该特征横向连接到下一层参考集上的特征，如此可以利用底层的定位细节信息。S202, feature pyramid network processing steps: as shown in Figure 3, the feature pyramid network performs top-down by combining high-level features with low resolution and high semantic information and low-level features with high resolution and low semantic information. The side connection makes the features at all scales have rich semantic information. This operation is performed simultaneously in the convolutional network instead of constructing an image pyramid, thereby avoiding additional time-consuming and memory usage. The convolutional network is a bottom-up convolution calculation. By defining a pyramid level for each network stage, the last layer output of each network stage is selected as the reference set of the feature map, and then each level is top-down The reference set is upsampled, and then the feature is horizontally connected to the feature on the next layer reference set, so that the underlying positioning details can be used.
通过卷积神经网络与特征金字塔网络的结合，可以提取到具有多尺度信息的图像特征图，该图像特征图作为输电线路部件自动识别步骤S3的输入数据。Through the combination of the convolutional neural network and the feature pyramid network, the image feature map with multi-scale information can be extracted, and the image feature map is used as the input data of the automatic identification step S3 of the transmission line components.
如图4、5所示，步骤S3中训练目标定位回归网络与分类网络步骤具体包括以下步骤：As shown in Figures 4 and 5, the steps of training the target positioning regression network and the classification network in step S3 specifically include the following steps:
S301、构建位置敏感分数图步骤：在卷积网络与特征金字塔网络的结合提取特征图后，添加一个卷积得分层用于提取图像上每个类别k2个位置敏感分数图，假设共有C+1个类别(C类目标，1类背景)，经过卷积产生k2(C+1)个通道的输出层。S301. Step of constructing a position-sensitive score map: After the feature map is extracted by combining the convolutional network and the feature pyramid network, a convolutional score layer is added to extract k2 position-sensitive score maps of each category on the image, assuming a total of C +1 category (C category target, 1 category background), after convolution to generate an output layer of k2 (C+1) channels.
S302、构建目标定位回归网络步骤：采用固定比例的候选框提取方式对特征图进行候选区域提取，固定比例分别为{1:2,1:1,2:1}。在步骤S2中使用了5种特征金字塔的级别，所以共获得15种区域提取框。为实现对目标定位回归网络的训练，定义损失函数为：S302. The step of constructing a target location regression network: extracting candidate regions from the feature map by using a fixed-ratio candidate frame extraction method, and the fixed ratios are respectively {1:2, 1:1, 2:1}. In step S2, 5 levels of the feature pyramid are used, so a total of 15 types of region extraction frames are obtained. In order to realize the training of the target localization regression network, the loss function is defined as:
其中，i是区域提取框的序号，pi是区域提取框时目标概率，是真实标签的概率：若是正样本时为1，是负样本时为0，正负样本采用区域提取框与真实边框的IoU判定。当IoU大于0.7时为正样本，IoU小于0.3时为负样本，而在0.3-0.7之间的全部忽略。ti是预测边框的位置坐标，是真实边框的位置坐标，Lcls是分类损失函数，用来判别边框里是否含有目标；Lreg是回归函数，对边框做位置和大小的微调，Ncls是训练时mini-batch的大小，大约为2000，Nreg是区域提取框定位的数量，λ是平衡参数，设置为15。Among them, i is the serial number of the region extraction frame, pi is the target probability when the region is extracted, Probability of being a true label: 1 if it is a positive sample, and 0 if it is a negative sample. The positive and negative samples are judged by the IoU of the area extraction frame and the real border. When the IoU is greater than 0.7, it is a positive sample, when the IoU is less than 0.3, it is a negative sample, and all the samples between 0.3-0.7 are ignored. ti is the position coordinate of the predicted bounding box, is the position coordinate of the real border, Lcls is the classification loss function, used to judge whether there is a target in the border; Lreg is a regression function, fine-tuning the position and size of the border, Ncls is the size of the mini-batch during training, about is 2000, Nreg is the number of region extraction frame locations, and λ is the balance parameter, which is set to 15.
为了实现对网络的整体训练，对卷积得分层、分类网络及目标定位回归网络进行参数随机初始化，定义损失函数为：In order to realize the overall training of the network, the parameters of the convolutional layer, classification network and target positioning regression network are randomly initialized, and the loss function is defined as:
其中，c*是候选区域的真实标签，[c*＞0]代表如果是真实标签时，参数设为1，否则为0；是用于分类的交叉熵损失函数，Lreg(t,t*)是边界回归损失函数，t*是真实标签的坐标，t是输出的回归坐标，平衡参数λ此处设为1。Among them, c* is the real label of the candidate area, [c* > 0] means that if it is a real label, the parameter is set to 1, otherwise it is 0; is the cross-entropy loss function for classification, Lreg (t,t* ) is the boundary regression loss function, t* is the coordinate of the real label, t is the regression coordinate of the output, and the balance parameter λ is set to 1 here.
S303、构建分类网络步骤：将感兴趣区域的标记框映射到卷积得分层上，并且划分k*k个分格，每个区域得到C+1个维度的特征图，按顺序依次将k*k个格子上对应的得分图抽取出来，然后这k2个得分图用来对候选区域进行投票从而判定目标的类别；S303, the step of constructing a classification network: map the marked frame of the region of interest to the convolutional score layer, and divide k*k grids, each region obtains a feature map of C+1 dimensions, and k in sequence *The corresponding score maps on the k grids are extracted, and then these k2 score maps are used to vote for the candidate area to determine the category of the target;
对于大小为w*h的感兴趣候选区域，每个分格的大小为(w/k)*(h/k)，对于第i行第j列的分格中的得分图，进行如下定义：For a candidate region of interest with a size of w*h, the size of each grid is (w/k)*(h/k), and for the score map in the grid of row i and column j, define as follows:
其中，rc(i,j)是针对第c类目标的第(i,j)个分格的响应得分，Zi,j,c代表k2(C+1)个得分图中的一个，(x0,y0)表示候选区域的左上角坐标，n表示每个格子中像素的个数，θ是网络中所有可学习的参数；Among them, rc (i, j) is the response score of the (i, j)th cell for the c-th category target, Zi, j, c represents one of the k2 (C+1) score maps, (x0 , y0 ) represents the coordinates of the upper left corner of the candidate area, n represents the number of pixels in each grid, and θ is all learnable parameters in the network;
对于k2个得分图通过平均得分进行投票，对于每个候选区域产生C+1维的向量：For k2 score maps, vote by the average score, and generate a C+1-dimensional vector for each candidate region:
最后计算每个目标的softmax响应：Finally compute the softmax response for each target:
用于输电线路巡检图像识别的特征提取网络、区域提取网络与分类网络，是一个顺序过程，在训练时采用端对端的策略进行训练。通过VGG-net16网络初始化特征提取网络参数，区域提取网络与分类网络权重参数采用高斯函数进行随之初始化，利用fine-tune策略对其进行优化，得到最终输电线路巡检图像识别模型。The feature extraction network, region extraction network and classification network used for transmission line inspection image recognition are a sequential process, and an end-to-end strategy is used for training. Through the VGG-net16 network initialization feature extraction network parameters, the region extraction network and classification network weight parameters are initialized with Gaussian functions, and the fine-tune strategy is used to optimize them to obtain the final transmission line inspection image recognition model.
如图6所示，输电线路部件检测步骤包括以下步骤:As shown in Figure 6, the transmission line component detection step includes the following steps:
S401、初始化网络步骤：利用优化、训练得到的图像识别模型初始化训练网络，包括网络结构定义、网络基本参数定义、权重及偏置数据填充；S401. Initialize the network step: use the optimized and trained image recognition model to initialize the training network, including network structure definition, network basic parameter definition, weight and bias data filling;
S402、提取图像特征图步骤：首先，对输入图像进行多层卷积操作提取输入图像的高层语意表达；其次，对提取到的特征进行多尺度分割与上采样；最后，将金字塔特征与卷积后的特征进行融合，形成最终特征图；S402. The step of extracting image feature maps: firstly, perform multi-layer convolution operation on the input image to extract the high-level semantic expression of the input image; secondly, perform multi-scale segmentation and up-sampling on the extracted features; finally, combine the pyramid features with convolution The final features are fused to form the final feature map;
S403、提取候选区域步骤：利用区域提取网络提取候选框，并进行非极大值抑制操作，只保留得分最高的前300个候选框，以实现检测的快速处理；S403, the step of extracting candidate regions: using the region extraction network to extract candidate frames, and performing a non-maximum value suppression operation, only retaining the top 300 candidate frames with the highest scores, so as to realize rapid processing of detection;
S404、计算位置敏感得分图步骤：对候选框进行得分图卷积操作，计算每个候选框的得分，并求取每个候选框的得分均值，构建位置敏感得分图；S404. The step of calculating a position-sensitive score map: perform a score map convolution operation on the candidate frame, calculate the score of each candidate frame, and obtain the mean value of the score of each candidate frame, and construct a position-sensitive score map;
S405、确定候选区域类别与标记框步骤：对敏感得分图进行softmax分类，确定最终类别；对候选框进行拟合优化确定最终目标标记位置。S405. The step of determining the category of the candidate area and the marked frame: perform softmax classification on the sensitive score map to determine the final category; perform fitting optimization on the candidate frame to determine the final target marked position.
上述虽然结合附图对本发明的具体实施方式进行了描述，但并非对本发明保护范围的限制，所属领域技术人员应该明白，在本发明的技术方案的基础上，本领域技术人员不需要付出创造性劳动即可做出的各种修改或变形仍在本发明的保护范围以内。Although the specific implementation of the present invention has been described above in conjunction with the accompanying drawings, it does not limit the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, those skilled in the art do not need to pay creative work Various modifications or variations that can be made are still within the protection scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710605607.6ACN107392901A (en) | 2017-07-24 | 2017-07-24 | A kind of method for transmission line part intelligence automatic identification |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710605607.6ACN107392901A (en) | 2017-07-24 | 2017-07-24 | A kind of method for transmission line part intelligence automatic identification |
| Publication Number | Publication Date |
|---|---|
| CN107392901Atrue CN107392901A (en) | 2017-11-24 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710605607.6APendingCN107392901A (en) | 2017-07-24 | 2017-07-24 | A kind of method for transmission line part intelligence automatic identification |
| Country | Link |
|---|---|
| CN (1) | CN107392901A (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107993228A (en)* | 2017-12-15 | 2018-05-04 | 中国人民解放军总医院 | A kind of vulnerable plaque automatic testing method and device based on cardiovascular OCT images |
| CN108010030A (en)* | 2018-01-24 | 2018-05-08 | 福州大学 | A kind of Aerial Images insulator real-time detection method based on deep learning |
| CN108009515A (en)* | 2017-12-14 | 2018-05-08 | 杭州远鉴信息科技有限公司 | A kind of power transmission line positioning identifying method of the unmanned plane image based on FCN |
| CN108038519A (en)* | 2018-01-30 | 2018-05-15 | 浙江大学 | A kind of uterine neck image processing method and device based on dense feature pyramid network |
| CN108032994A (en)* | 2017-12-06 | 2018-05-15 | 余姚市荣事特电子有限公司 | A kind of unmanned helicopter |
| CN108229587A (en)* | 2018-02-06 | 2018-06-29 | 贵州电网有限责任公司 | A kind of autonomous scan method of transmission tower based on aircraft floating state |
| CN108280855A (en)* | 2018-01-13 | 2018-07-13 | 福州大学 | A kind of insulator breakdown detection method based on Fast R-CNN |
| CN108647655A (en)* | 2018-05-16 | 2018-10-12 | 北京工业大学 | Low latitude aerial images power line foreign matter detecting method based on light-duty convolutional neural networks |
| CN108693441A (en)* | 2018-04-16 | 2018-10-23 | 华北电力大学（保定） | A kind of electric transmission line isolator recognition methods and system |
| CN108734143A (en)* | 2018-05-28 | 2018-11-02 | 江苏迪伦智能科技有限公司 | A kind of transmission line of electricity online test method based on binocular vision of crusing robot |
| CN108764235A (en)* | 2018-05-23 | 2018-11-06 | 中国民用航空总局第二研究所 | Neural network model, object detection method, equipment and medium |
| CN108875819A (en)* | 2018-06-08 | 2018-11-23 | 浙江大学 | A kind of object and component associated detecting method based on shot and long term memory network |
| CN108898177A (en)* | 2018-06-26 | 2018-11-27 | 广东电网有限责任公司 | A kind of power transmission line corridor earth surface classification method, system, medium and equipment |
| CN109035204A (en)* | 2018-06-25 | 2018-12-18 | 华南理工大学 | A kind of weld seam target real-time detection method |
| CN109102502A (en)* | 2018-08-03 | 2018-12-28 | 西北工业大学 | Pulmonary nodule detection method based on Three dimensional convolution neural network |
| CN109376768A (en)* | 2018-09-21 | 2019-02-22 | 福州大学 | A deep learning-based fault diagnosis method for aerial image tower signage |
| CN109377483A (en)* | 2018-09-30 | 2019-02-22 | 云南电网有限责任公司普洱供电局 | Porcelain insulator crack detecting method and device |
| CN109409503A (en)* | 2018-09-27 | 2019-03-01 | 深圳市铱硙医疗科技有限公司 | Training method, image conversion method, device, equipment and the medium of neural network |
| CN109472214A (en)* | 2018-10-17 | 2019-03-15 | 福州大学 | A real-time detection method of aerial foreign object images based on deep learning |
| CN109544501A (en)* | 2018-03-22 | 2019-03-29 | 广东电网有限责任公司清远供电局 | A kind of transmission facility defect inspection method based on unmanned plane multi-source image characteristic matching |
| CN109614488A (en)* | 2018-12-04 | 2019-04-12 | 广西大学 | Discrimination method of live working conditions of distribution network based on text classification and image recognition |
| CN109657677A (en)* | 2018-12-14 | 2019-04-19 | 广东电网有限责任公司 | A kind of detection method of electric power line pole tower, device, equipment and storage medium |
| CN109711437A (en)* | 2018-12-06 | 2019-05-03 | 武汉三江中电科技有限责任公司 | A kind of transformer part recognition methods based on YOLO network model |
| CN109859163A (en)* | 2018-12-19 | 2019-06-07 | 重庆邮电大学 | A kind of LCD defect inspection method based on feature pyramid convolutional neural networks |
| CN109885091A (en)* | 2019-03-21 | 2019-06-14 | 华北电力大学（保定） | A method and system for autonomous flight control of an unmanned aerial vehicle |
| CN109977963A (en)* | 2019-04-10 | 2019-07-05 | 京东方科技集团股份有限公司 | Image processing method, unit and computer-readable medium |
| CN110070124A (en)* | 2019-04-15 | 2019-07-30 | 广州小鹏汽车科技有限公司 | A kind of image amplification method and system based on production confrontation network |
| CN110070530A (en)* | 2019-04-19 | 2019-07-30 | 山东大学 | A kind of powerline ice-covering detection method based on deep neural network |
| CN110085218A (en)* | 2019-03-26 | 2019-08-02 | 天津大学 | A kind of audio scene recognition method based on feature pyramid network |
| CN110097090A (en)* | 2019-04-10 | 2019-08-06 | 东南大学 | A kind of image fine granularity recognition methods based on multi-scale feature fusion |
| CN110136097A (en)* | 2019-04-10 | 2019-08-16 | 南方电网科学研究院有限责任公司 | Insulator fault identification method and device based on characteristic pyramid |
| CN110443130A (en)* | 2019-07-01 | 2019-11-12 | 国网湖南省电力有限公司 | A kind of electric distribution network overhead wire abnormal state detection method |
| CN111144420A (en)* | 2019-11-19 | 2020-05-12 | 国网通用航空有限公司 | Small sample-based electric tower defect monitoring method and system |
| WO2020093624A1 (en)* | 2018-11-07 | 2020-05-14 | 五邑大学 | Antenna downward inclination angle measurement method based on multi-scale detection algorithm |
| CN111160293A (en)* | 2019-12-31 | 2020-05-15 | 珠海大横琴科技发展有限公司 | Small target ship detection method and system based on characteristic pyramid network |
| CN111179262A (en)* | 2020-01-02 | 2020-05-19 | 国家电网有限公司 | Electric power inspection image hardware fitting detection method combined with shape attribute |
| CN111275015A (en)* | 2020-02-28 | 2020-06-12 | 广东电网有限责任公司 | Unmanned aerial vehicle-based power line inspection electric tower detection and identification method and system |
| CN111291739A (en)* | 2020-05-09 | 2020-06-16 | 腾讯科技（深圳）有限公司 | Face detection, image detection neural network training method, device and equipment |
| CN111340189A (en)* | 2020-02-21 | 2020-06-26 | 之江实验室 | A Realization Method of Spatial Pyramid Graph Convolutional Network |
| CN111462310A (en)* | 2020-03-31 | 2020-07-28 | 国网辽宁省电力有限公司朝阳供电公司 | A spatial location method of bolt defects based on multi-view geometry |
| CN111597964A (en)* | 2020-05-13 | 2020-08-28 | 云南电网有限责任公司电力科学研究院 | A kind of insulator image detection method |
| CN111738188A (en)* | 2020-06-29 | 2020-10-02 | 广东电网有限责任公司 | Power transmission line crimping hardware identification method and device, terminal and storage medium |
| CN111797890A (en)* | 2020-05-18 | 2020-10-20 | 中国电力科学研究院有限公司 | A method and system for detecting defects in transmission line equipment |
| CN112001902A (en)* | 2020-08-19 | 2020-11-27 | 上海商汤智能科技有限公司 | Defect detection method and related device, equipment and storage medium |
| CN112307851A (en)* | 2019-08-02 | 2021-02-02 | 上海交通大学烟台信息技术研究院 | A method and system for identifying a bird's nest on a power tower |
| CN112380944A (en)* | 2020-11-06 | 2021-02-19 | 中国电力科学研究院有限公司 | Method and system for evaluating structural state of transmission tower |
| CN112528971A (en)* | 2021-02-07 | 2021-03-19 | 北京智芯微电子科技有限公司 | Power transmission line abnormal target detection method and system based on deep learning |
| CN112669282A (en)* | 2020-12-29 | 2021-04-16 | 燕山大学 | Spine positioning method based on deep neural network |
| CN112906811A (en)* | 2021-03-09 | 2021-06-04 | 西安电子科技大学 | Automatic classification method for images of engineering vehicle-mounted equipment based on Internet of things architecture |
| CN115482214A (en)* | 2022-09-16 | 2022-12-16 | 西安热工研究院有限公司 | Photovoltaic power station component quantity measuring method based on unmanned aerial vehicle collected image |
| CN115546568A (en)* | 2022-12-01 | 2022-12-30 | 合肥中科类脑智能技术有限公司 | Insulator defect detection method, system, equipment and storage medium |
| CN116091428A (en)* | 2022-12-29 | 2023-05-09 | 国网电力空间技术有限公司 | High-precision intelligent power transmission line inspection image tower dividing method and system |
| CN116468998A (en)* | 2022-09-09 | 2023-07-21 | 国网湖北省电力有限公司超高压公司 | Visual characteristic-based power transmission line small part and hanging point part detection method |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106504233A (en)* | 2016-10-18 | 2017-03-15 | 国网山东省电力公司电力科学研究院 | Image electric power widget recognition methodss and system are patrolled and examined based on the unmanned plane of Faster R CNN |
| CN106874894A (en)* | 2017-03-28 | 2017-06-20 | 电子科技大学 | A kind of human body target detection method based on the full convolutional neural networks in region |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106504233A (en)* | 2016-10-18 | 2017-03-15 | 国网山东省电力公司电力科学研究院 | Image electric power widget recognition methodss and system are patrolled and examined based on the unmanned plane of Faster R CNN |
| CN106874894A (en)* | 2017-03-28 | 2017-06-20 | 电子科技大学 | A kind of human body target detection method based on the full convolutional neural networks in region |
| Title |
|---|
| JIFENG DAI等: "R-FCN: Object Detection via Region-based Fully Convolutional Networks", 《COMPUTER VISION AND PATTERN RECOGNITION》* |
| SHAOQING REN等: "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108032994A (en)* | 2017-12-06 | 2018-05-15 | 余姚市荣事特电子有限公司 | A kind of unmanned helicopter |
| CN108009515A (en)* | 2017-12-14 | 2018-05-08 | 杭州远鉴信息科技有限公司 | A kind of power transmission line positioning identifying method of the unmanned plane image based on FCN |
| CN108009515B (en)* | 2017-12-14 | 2022-04-22 | 杭州远鉴信息科技有限公司 | Power transmission line positioning and identifying method of unmanned aerial vehicle aerial image based on FCN |
| CN107993228A (en)* | 2017-12-15 | 2018-05-04 | 中国人民解放军总医院 | A kind of vulnerable plaque automatic testing method and device based on cardiovascular OCT images |
| CN108280855A (en)* | 2018-01-13 | 2018-07-13 | 福州大学 | A kind of insulator breakdown detection method based on Fast R-CNN |
| CN108010030A (en)* | 2018-01-24 | 2018-05-08 | 福州大学 | A kind of Aerial Images insulator real-time detection method based on deep learning |
| CN108038519A (en)* | 2018-01-30 | 2018-05-15 | 浙江大学 | A kind of uterine neck image processing method and device based on dense feature pyramid network |
| CN108038519B (en)* | 2018-01-30 | 2020-11-24 | 浙江大学 | A cervical image processing method and device based on dense feature pyramid network |
| CN108229587B (en)* | 2018-02-06 | 2021-04-06 | 贵州电网有限责任公司 | Autonomous transmission tower scanning method based on hovering state of aircraft |
| CN108229587A (en)* | 2018-02-06 | 2018-06-29 | 贵州电网有限责任公司 | A kind of autonomous scan method of transmission tower based on aircraft floating state |
| CN109544501A (en)* | 2018-03-22 | 2019-03-29 | 广东电网有限责任公司清远供电局 | A kind of transmission facility defect inspection method based on unmanned plane multi-source image characteristic matching |
| CN108693441A (en)* | 2018-04-16 | 2018-10-23 | 华北电力大学（保定） | A kind of electric transmission line isolator recognition methods and system |
| CN108693441B (en)* | 2018-04-16 | 2021-01-22 | 华北电力大学（保定） | Method and system for identifying insulator of power transmission line |
| CN108647655A (en)* | 2018-05-16 | 2018-10-12 | 北京工业大学 | Low latitude aerial images power line foreign matter detecting method based on light-duty convolutional neural networks |
| CN108647655B (en)* | 2018-05-16 | 2022-07-12 | 北京工业大学 | Low-altitude aerial image power line foreign object detection method based on light convolutional neural network |
| CN108764235B (en)* | 2018-05-23 | 2021-06-29 | 中国民用航空总局第二研究所 | Object detection method, equipment and medium |
| CN108764235A (en)* | 2018-05-23 | 2018-11-06 | 中国民用航空总局第二研究所 | Neural network model, object detection method, equipment and medium |
| CN108734143A (en)* | 2018-05-28 | 2018-11-02 | 江苏迪伦智能科技有限公司 | A kind of transmission line of electricity online test method based on binocular vision of crusing robot |
| CN108875819B (en)* | 2018-06-08 | 2020-10-27 | 浙江大学 | Object and component joint detection method based on long-term and short-term memory network |
| CN108875819A (en)* | 2018-06-08 | 2018-11-23 | 浙江大学 | A kind of object and component associated detecting method based on shot and long term memory network |
| CN109035204A (en)* | 2018-06-25 | 2018-12-18 | 华南理工大学 | A kind of weld seam target real-time detection method |
| CN109035204B (en)* | 2018-06-25 | 2021-06-08 | 华南理工大学 | Real-time detection method for weld joint target |
| CN108898177A (en)* | 2018-06-26 | 2018-11-27 | 广东电网有限责任公司 | A kind of power transmission line corridor earth surface classification method, system, medium and equipment |
| CN109102502B (en)* | 2018-08-03 | 2021-07-23 | 西北工业大学 | Pulmonary nodule detection method based on three-dimensional convolutional neural network |
| CN109102502A (en)* | 2018-08-03 | 2018-12-28 | 西北工业大学 | Pulmonary nodule detection method based on Three dimensional convolution neural network |
| CN109376768B (en)* | 2018-09-21 | 2021-12-17 | 福州大学 | Aerial image tower signboard fault diagnosis method based on deep learning |
| CN109376768A (en)* | 2018-09-21 | 2019-02-22 | 福州大学 | A deep learning-based fault diagnosis method for aerial image tower signage |
| CN109409503A (en)* | 2018-09-27 | 2019-03-01 | 深圳市铱硙医疗科技有限公司 | Training method, image conversion method, device, equipment and the medium of neural network |
| CN109377483A (en)* | 2018-09-30 | 2019-02-22 | 云南电网有限责任公司普洱供电局 | Porcelain insulator crack detecting method and device |
| CN109472214A (en)* | 2018-10-17 | 2019-03-15 | 福州大学 | A real-time detection method of aerial foreign object images based on deep learning |
| WO2020093624A1 (en)* | 2018-11-07 | 2020-05-14 | 五邑大学 | Antenna downward inclination angle measurement method based on multi-scale detection algorithm |
| US11145089B2 (en) | 2018-11-07 | 2021-10-12 | Wuyi University | Method for measuring antenna downtilt based on multi-scale detection algorithm |
| CN109614488A (en)* | 2018-12-04 | 2019-04-12 | 广西大学 | Discrimination method of live working conditions of distribution network based on text classification and image recognition |
| CN109711437A (en)* | 2018-12-06 | 2019-05-03 | 武汉三江中电科技有限责任公司 | A kind of transformer part recognition methods based on YOLO network model |
| CN109657677A (en)* | 2018-12-14 | 2019-04-19 | 广东电网有限责任公司 | A kind of detection method of electric power line pole tower, device, equipment and storage medium |
| CN109859163A (en)* | 2018-12-19 | 2019-06-07 | 重庆邮电大学 | A kind of LCD defect inspection method based on feature pyramid convolutional neural networks |
| CN109885091A (en)* | 2019-03-21 | 2019-06-14 | 华北电力大学（保定） | A method and system for autonomous flight control of an unmanned aerial vehicle |
| CN110085218A (en)* | 2019-03-26 | 2019-08-02 | 天津大学 | A kind of audio scene recognition method based on feature pyramid network |
| CN110097090A (en)* | 2019-04-10 | 2019-08-06 | 东南大学 | A kind of image fine granularity recognition methods based on multi-scale feature fusion |
| CN109977963B (en)* | 2019-04-10 | 2021-10-15 | 京东方科技集团股份有限公司 | Image processing method, apparatus, apparatus, and computer-readable medium |
| CN109977963A (en)* | 2019-04-10 | 2019-07-05 | 京东方科技集团股份有限公司 | Image processing method, unit and computer-readable medium |
| CN110136097A (en)* | 2019-04-10 | 2019-08-16 | 南方电网科学研究院有限责任公司 | Insulator fault identification method and device based on characteristic pyramid |
| CN110070124A (en)* | 2019-04-15 | 2019-07-30 | 广州小鹏汽车科技有限公司 | A kind of image amplification method and system based on production confrontation network |
| CN110070530A (en)* | 2019-04-19 | 2019-07-30 | 山东大学 | A kind of powerline ice-covering detection method based on deep neural network |
| CN110443130A (en)* | 2019-07-01 | 2019-11-12 | 国网湖南省电力有限公司 | A kind of electric distribution network overhead wire abnormal state detection method |
| CN112307851A (en)* | 2019-08-02 | 2021-02-02 | 上海交通大学烟台信息技术研究院 | A method and system for identifying a bird's nest on a power tower |
| CN111144420B (en)* | 2019-11-19 | 2023-08-11 | 国网电力空间技术有限公司 | Method and system for monitoring defects of electric towers based on small samples |
| CN111144420A (en)* | 2019-11-19 | 2020-05-12 | 国网通用航空有限公司 | Small sample-based electric tower defect monitoring method and system |
| CN111160293A (en)* | 2019-12-31 | 2020-05-15 | 珠海大横琴科技发展有限公司 | Small target ship detection method and system based on characteristic pyramid network |
| CN111179262A (en)* | 2020-01-02 | 2020-05-19 | 国家电网有限公司 | Electric power inspection image hardware fitting detection method combined with shape attribute |
| CN111179262B (en)* | 2020-01-02 | 2024-09-06 | 国家电网有限公司 | Electric power inspection image hardware fitting detection method combining shape attribute |
| CN111340189B (en)* | 2020-02-21 | 2023-11-24 | 之江实验室 | An implementation method of spatial pyramid graph convolution network |
| CN111340189A (en)* | 2020-02-21 | 2020-06-26 | 之江实验室 | A Realization Method of Spatial Pyramid Graph Convolutional Network |
| CN111275015A (en)* | 2020-02-28 | 2020-06-12 | 广东电网有限责任公司 | Unmanned aerial vehicle-based power line inspection electric tower detection and identification method and system |
| CN111462310B (en)* | 2020-03-31 | 2023-10-17 | 国网辽宁省电力有限公司朝阳供电公司 | Bolt defect space positioning method based on multi-view geometry |
| CN111462310A (en)* | 2020-03-31 | 2020-07-28 | 国网辽宁省电力有限公司朝阳供电公司 | A spatial location method of bolt defects based on multi-view geometry |
| CN111291739B (en)* | 2020-05-09 | 2020-09-18 | 腾讯科技（深圳）有限公司 | Face detection and image detection neural network training method, device and equipment |
| US12131580B2 (en) | 2020-05-09 | 2024-10-29 | Tencent Technology (Shenzhen) Company Limited | Face detection method, apparatus, and device, and training method, apparatus, and device for image detection neural network |
| CN111291739A (en)* | 2020-05-09 | 2020-06-16 | 腾讯科技（深圳）有限公司 | Face detection, image detection neural network training method, device and equipment |
| CN111597964B (en)* | 2020-05-13 | 2023-01-20 | 云南电网有限责任公司电力科学研究院 | Insulator image detection method |
| CN111597964A (en)* | 2020-05-13 | 2020-08-28 | 云南电网有限责任公司电力科学研究院 | A kind of insulator image detection method |
| CN111797890A (en)* | 2020-05-18 | 2020-10-20 | 中国电力科学研究院有限公司 | A method and system for detecting defects in transmission line equipment |
| CN111738188A (en)* | 2020-06-29 | 2020-10-02 | 广东电网有限责任公司 | Power transmission line crimping hardware identification method and device, terminal and storage medium |
| CN112001902A (en)* | 2020-08-19 | 2020-11-27 | 上海商汤智能科技有限公司 | Defect detection method and related device, equipment and storage medium |
| CN112380944B (en)* | 2020-11-06 | 2021-12-21 | 中国电力科学研究院有限公司 | A method and system for evaluating the structural state of transmission towers based on satellite remote sensing |
| CN112380944A (en)* | 2020-11-06 | 2021-02-19 | 中国电力科学研究院有限公司 | Method and system for evaluating structural state of transmission tower |
| CN112669282B (en)* | 2020-12-29 | 2023-02-14 | 燕山大学 | A spinal localization method based on deep neural network |
| CN112669282A (en)* | 2020-12-29 | 2021-04-16 | 燕山大学 | Spine positioning method based on deep neural network |
| CN112528971A (en)* | 2021-02-07 | 2021-03-19 | 北京智芯微电子科技有限公司 | Power transmission line abnormal target detection method and system based on deep learning |
| CN112528971B (en)* | 2021-02-07 | 2021-06-04 | 北京智芯微电子科技有限公司 | Method and system for abnormal target detection of transmission line based on deep learning |
| CN112906811A (en)* | 2021-03-09 | 2021-06-04 | 西安电子科技大学 | Automatic classification method for images of engineering vehicle-mounted equipment based on Internet of things architecture |
| CN112906811B (en)* | 2021-03-09 | 2023-04-18 | 西安电子科技大学 | Automatic classification method for images of engineering vehicle-mounted equipment based on Internet of things architecture |
| CN116468998A (en)* | 2022-09-09 | 2023-07-21 | 国网湖北省电力有限公司超高压公司 | Visual characteristic-based power transmission line small part and hanging point part detection method |
| CN115482214A (en)* | 2022-09-16 | 2022-12-16 | 西安热工研究院有限公司 | Photovoltaic power station component quantity measuring method based on unmanned aerial vehicle collected image |
| CN115546568A (en)* | 2022-12-01 | 2022-12-30 | 合肥中科类脑智能技术有限公司 | Insulator defect detection method, system, equipment and storage medium |
| CN116091428A (en)* | 2022-12-29 | 2023-05-09 | 国网电力空间技术有限公司 | High-precision intelligent power transmission line inspection image tower dividing method and system |
| CN116091428B (en)* | 2022-12-29 | 2023-09-01 | 国网电力空间技术有限公司 | High-precision intelligent power transmission line inspection image tower dividing method and system |
| Publication | Publication Date | Title |
|---|---|---|
| CN107392901A (en) | A kind of method for transmission line part intelligence automatic identification | |
| CN113378686B (en) | Two-stage remote sensing target detection method based on target center point estimation | |
| CN113076816B (en) | Hot spot recognition method for solar photovoltaic modules based on infrared and visible light images | |
| CN108734143A (en) | A kind of transmission line of electricity online test method based on binocular vision of crusing robot | |
| CN113920107A (en) | A method of insulator damage detection based on improved yolov5 algorithm | |
| CN111179217A (en) | A multi-scale target detection method in remote sensing images based on attention mechanism | |
| CN110533631A (en) | SAR image change detection based on the twin network of pyramid pondization | |
| CN112164038A (en) | Photovoltaic hot spot detection method based on deep convolutional neural network | |
| CN109614985A (en) | A target detection method based on densely connected feature pyramid network | |
| CN109584248A (en) | Infrared surface object instance dividing method based on Fusion Features and dense connection network | |
| CN114565824B (en) | Single-stage rotating ship detection method based on full convolution network | |
| CN106127204A (en) | A kind of multi-direction meter reading Region detection algorithms of full convolutional neural networks | |
| CN111444939A (en) | Small-scale equipment component detection method based on weakly supervised collaborative learning in open scenarios in the power field | |
| CN115457006A (en) | Method and device for UAV inspection defect classification based on similarity and consistency self-distillation | |
| CN107463881A (en) | A kind of character image searching method based on depth enhancing study | |
| Li et al. | Learning to holistically detect bridges from large-size VHR remote sensing imagery | |
| CN113487600B (en) | Feature enhancement scale self-adaptive perception ship detection method | |
| CN113361528B (en) | Multi-scale target detection method and system | |
| CN113298767A (en) | Reliable go map recognition method capable of overcoming light reflection phenomenon | |
| CN111414855B (en) | Telegraph pole sign target detection and identification method based on end-to-end regression model | |
| CN114926718A (en) | Low-small slow target detection method with fusion of adjacent scale weight distribution characteristics | |
| CN113269171B (en) | Lane line detection method, electronic device and vehicle | |
| CN112084860A (en) | Target object detection method and device and thermal power plant detection method and device | |
| CN117392111A (en) | Network and method for detecting surface defects of strip steel camouflage | |
| CN115410024A (en) | Power image defect detection method based on dynamic activation thermodynamic diagram |
| 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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information | Address after:250001 No. three, No. 17, Shizhong Road, Shizhong District, Shandong, Ji'nan Applicant after:State Grid Shandong Electric Power Company Information Communication Company Applicant after:National Network Intelligent Technology Co., Ltd. Applicant after:State Grid Corporation of China Address before:250001 No. three, No. 17, Shizhong Road, Shizhong District, Shandong, Ji'nan Applicant before:State Grid Shandong Electric Power Company Information Communication Company Applicant before:Shandong Luneng Intelligent Technology Co., Ltd. Applicant before:State Grid Corporation of China | |
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication | Application publication date:20171124 |