Abstract: To improve the safety of electrified railway operation and realize the automatic detection of catenary insulator defects, a detection method using edge pixel difference of adjacent insulator pieces based on deep convolutional neural network precise positioning is proposed for detecting insulator defects. The proposed method is used to detect the insulator defects on 200 catenary images for testing.Results show that the effective segmentation ratio of insulators is 98.5%,and the misidentification rate of defective insulators is 5.3%,proving that the method can effectively carrying out automatic detection of catenary insulator defects.