| Transmission lines exposed to the natural environment are prone to various problems such as aging,damage,oxidation and corrosion,which pose a potential threat to power transmission.Therefore,it is of great significance to pay attention to the safety of transmission lines.Traditional transmission line inspection methods are mostly manual.This method has a large workload and low efficiency,and it has certain dangers.In response to the above problems,it was decided to use a drone instead of a manual inspection.In the whole process of drone inspection of power lines,the key issue is the detection and fault classification identification of power lines.This paper proposes a method for detecting and classifying faults in power transmission lines.The main work is as follows:1.Pre-processing operations on the aerial image of the transmission line,including graying,filtering,and histogram equalization.Perform edge detection on the preprocessed image to extract image edge features.In the edge detection process,the Edge Drawing algorithm is optimized,and then compared with the commonly used detection algorithms such as Canny algorithm and Scharr filter.2.Clustering and linear sensing grouping of image pixels after edge detection,eliminating pixels on the line,and using random Hough transform to identify the power line.3.Construct a power line image dataset and use the convolutional neural network--Inception v3 network,to train a model for the linear dataset.Then combined with the idea of migration learning,using the model migration method,the model of the training line is transferred to the training power line data set.During the training process,the activation function replaces the Relu function with the P-Relu function to solve the dead zone problem of the Relu function.A Softmax layer is added to the output layer to classify transmission line faults.The transmission line detection and fault classification method proposed in thispaper can quickly detect the transmission line,and the fault classification accuracy is better than other algorithms,indicating that the proposed method is feasible. |
