Research on Change Detection Technology of Equipment Appearance Defect in Substation Scene

With the continuous development of new technologies such as big data, artificial intelligence and cloud computing, intelligent terminal devices such as patrol robots, high-definition videos and unmanned aerial vehicles have been applied injoint patrol inspection of substations. In the face of the demand for appearance defect detection in the process of joint patrol inspection of substations, this paper introduces a twin differential attention change detection model (Diff-SANet) proposed by the China Electric Power Research Institute, which is used to solve the problem of defect status determination and defect area location when the equipment appearance changes to produce defects, and the defect types cannot be pre-defined and cannot be fully exemplified before the occurrence of defects. Diff -SANet network model can only use the normal state image set of observation points to complete model training, and the trained model can be used to determine appearance defects and locate areas. Finally, this paper verifies the real data set of the substation, and the experimental results show that the accuracy of the model can reach 91.6%. This method provides guidance for the promotion of intelligent patrol equipment application in substations.