Research On Electric Power Equipment Image Classification Based On Deep Learning

With the expansion of the grid and the construction of smart grids,video surveillance and image acquisition technologies are widely used in the field of power systems.However,due to the variety of power equipment and the complexity of the field environment,it is a difficult research point to intelligently process the collected electric power equipment images and automatically detect and classify the target equipment contained therein.Aiming at the above problems,this paper applies image processing and deep learning technology to electric power equipment images,and studies the image extraction and classification methods of power equipment based on deep learning,which is of great significance to improve the intelligent level of power systems.Firstly,the feature analysis of the collected power equipment images is carried out.According to the characteristics of multiple devices and complex background in the equipment images,we analyze the feasibility of using YOLO network model to classify the target equipment in the images.Next,according to some shortcomings of YOLO network,this paper puts forward some improvement measures.The improved K-means algorithm is used to select the initial bounding box.so that it can fit the actual object faster in the training process and speed up the training of the network.By modifying the network structure of YOLO,the high-level semantic features and the low-level details are combined to improve the ability of the network to extract detailed features,which is more conducive to the detection and classification of small objects.Modifying the loss function can optimize the smaller object equally with the larger object in the training process.In addition,according to the characteristics of the image of cantilever knife-gate equipment,the concept of position confidence is proposed and the prior position information is introduced into the process of detection and classification.Finally,based on the convolutional neural network,an image classification model of electric power equipment based on improved YOLO network is designed.And the training of the model is completed and related experiments are carried out.Through the analysis of the experimental results,we can draw the conclusion that the improved YOLO-M network model has better classification performance,higher accuracy and recall rate than the original network.By comparing with other detection methods,it further proves that the network has good detection performance for small target devices.