Insulator Image Defect Detection Algorithm Based on Improved YOLOv7

The normal power supply of the transmission system provides people with life security, and the insulator is an important component of the transmission system, and the defects of the insulator will seriously affect the normal operation of the power transmission system. Therefore, it is necessary to detect whether the insulator is in good condition in time to avoid risks. Traditional detection methods are prone to problems such as missed detection, false detection, low efficiency and low accuracy in identifying insulators. In this paper, the improved algorithms such as DSConv distributed offset convolution, SimAM attention mechanism, MPDIoU loss function, and CARAFE upsampling operator are added to the YOLOv7 model, which is based on the YOLOv7 model, and the DCSA-YOLOv7 model for detecting insulator defects is formed. The improved model accuracy reached 98.9%, an increase of 1.8% compared to YOLOv7 accuracy, and a 1.6% increase in recall. The experimental results show that the method adopted in this paper can detect insulator defects more accurately.