Hardware Detection Method Of Transmission Line Patrol Inspection Image Based On Improved Yolov4 Model

In order to solve the problem of intelligent hardware detection in aerial images, a hardware target detection method based on improved YOLOV4 model is proposed. In order to solve the problems of dense hardware and occlusion in aerial images, the improved network based on channel and spatial hybrid attention mechanism can further improve the detection effect of dense occlusion hardware and reduce image false detection and missed detection. In order to solve the problem that there is a great error in the position of the detection frame caused by the interference between the hardware and the hardware and between the hardware and the background, the prior frame is optimized by K-means++, and it is determined that the anchors generated by K=12 is the best, and the detection boxes are more suitable for the target. The experimental results show that the proposed method solves the problems of missing detection, misdetection and inaccurate detection frame to some extent, in which the mAP (mean Average Precision) value of the performance index is increased from 65.03% to 70.72%. The research can lay a good foundation for further state detection and fault diagnosis of typical hardware.