Research On Recognition And Location For Power Line Fault Based On Digital Image

It is very important for power line inspection to ensure the security and stability of power supply,and the main means of power line inspection include the helicopter,UAV etc.A large number of image data can be obtained during the inspection process,and the digital image processing technology is used to deal with it.Because the traditional detection algorithm depends on the prior knowledge of the developers to the image content,so the well performance is achieved only in specific scenario,and the universality and applicability of it is generally poor.Deep learning can automatically extract image features,and form a distributed representation of features,which is more conducive to recognition and classification.Moreover deep learning algorithm also has a good generalization ability.In order to improve the detection performance and the applicability of detection algorithm,this paper proposes a new algorithm to identify and locate the faults of power lines by using the convolution neural network,a deep learning algorithm.In this paper,a convolution neural network with six layers structure is proposed for object recognition,and the location algorithm based on the output map information is presented according to the characteristics of the power line.Firstly,the convolution neural network is used to identify each windows of the input image by the sliding window method and the output image is obtained.And then the output map is preprocessed by the binary operation and open operation.At last,the corresponding object is located according to the output map information.In the course of the network training,the data augmentation is used to increase the number of data,and the training set consists of 17000 image patch and the test set consists of 4000 image patch are obtained for the training of the network.The correct rate of the trained CNN network is 88.68% in the test set.In the 125 location detection pictures,there are 91.2% pictures whose power line and its faults can be accurately positioned.The experimental results show that the deep learning algorithm can accurately identify and locate power lines and its faults in images with complex background,and improve the generality and applicability of the detection algorithm.It has practical significance to improve the fault detection efficiency of the power line and to improve the detection accuracy.