Research on the Missing Cap Fault Detection Technology of Transmission Line Insulator Based on Deep Learning

In order to obtain more accurate and general detection performance, based on the feature pyramid structure and local attention mechanism, this paper proposes an improved algorithm LAPF based on Faster R-CNN to improve the detection accuracy of insulators and defects. Firstly, aiming at the problems of low detection accuracy caused by blurred insulator edges, small defect targets, and easy loss of details during detection, the LAPF detection model proposed in this paper introduces feature pyramid network and adds local attention mechanism. Combining bottom-up and top-down methods to obtain strong semantic features and enhance the learning of the underlying details, so as to improve the detection accuracy of transmission line insulator defect detection model. Secondly, this paper adds a local attention mechanism on the basis of FPN structure to extract feature information more accurately to alleviate the difficulty of small target detection, so as to improve the detection accuracy of small targets such as insulator defects of transmission lines. The experimental results of this paper show that the detection accuracy of LAPF detection model on the data set reaches 0.924, which fully demonstrates that the proposed improved model LAPF and target fusion data enhancement method are effective and necessary for realizing more accurate and general insulator defect detection in smart grid.