Defect Detection Of Electrical Circuit Based On Aerial Image

The defects of electrical circuit will cause short circuit and large area power failure,which will affect the safe operation of power grid.At present,the common way to ensure the normal operation of the electrical circuit is the helicopter or unmanned aerial vehicle inspection,which can obtain the inspection image through the camera and be processed by the staff.With the increase of the route inspection,the workload of power grid workers will increase accordingly,and the problem of leakage and misjudgment will easily occur.In this paper faults are identified in two practical production and operation environments of transmission line and distribution line respectively.The ceramic insulator of distribution line is prone to pollution flashover in humid environment and thunderstorm weather.In this paper,a method to identify contaminated insulators of distribution lines based on statistical characteristics is proposed.Based on the different color between insulator and background,the insulator is extracted by combining color features.Then the statistical features of the insulator images are extracted and used to train the classifier.Many transmission lines are built in the wilderness.Insulators are prone to self-explosion due to temperature changes or external forces.Birds nesting in towers can cause short circuit.Aiming at above-mentioned fault problem of transmission line,this paper realizes the classification and detection of bird's nest and fault insulator by building convolutional neural network,and solves the problem of weak generalization ability and low accuracy of traditional algorithm.Firstly,an appropriate network model is constructed by studying the convolution neural network.Then the classifier is trained and optimized in the course of training according to relevant training methods.Finally,the classifier is compared with the traditional feature extraction method.The experimental results show that the average accuracy rate of convolutional neural network is 10% higher than the traditional method for the bird's nest and fault insulator,that is,the performance classification effect of convolutional neural network is better than the traditional feature extraction algorithm.In this paper,the target detection algorithm is used to train the detector to detect the bird nest and the fault insulator in aerial images.Aiming at the problem of leakage detection both in classifier and detector,the method of cascading classifier and detector is adopted.The experimental results show that the miss rate of cascade network is 0,that is,staff do not need to pay attention to the images that are judged as fault-free by cascade network,so the fault identification can be completed well,which greatly reduces the labor intensity of the staff.