基于 EPSA-YOLOv5电力高空作业安全带佩戴检测.

To address the problem of missed detection and slow detection speed in safety belt wearing test for electric aloft work, this paper proposed a method for detecting the wearing of safety belts based on EPSA-YOLOv5 algorithm. This method was based on EPSANet backbone feature extraction network, which reduced the number of parameters in the network while maintaining good feature extraction performance, and speeding up the model recognition speed. By improving the spatial pyramid pooling structure, the model detection accuracy was improved; on this basis, an improved algorithm based on Soft-NMS was proposed to reduce the detection of targets. Experimental results show that the detection accuracy and speed of safety belt for aloft work based on EPSA-YOLOv5 network model are 2.34% higher than that of the original YOLOv5 model, which has practicality and efficiency. [ABSTRACT FROM AUTHOR]