Foreign objects on power lines especially on high-voltage power lines can threaten transmission safety. Therefore inspection of power lines is very important. In the inspection process, people often face low-illumination situations, which could cause great difficulties in the inspection. In this paper, I conducted this study to use the improved Retinex based on HSV color space for image enhancement and then use the improved YOLOv5 with SimAM attention mechanisms for power line foreign object detection. Evaluation of the algorithm through multiple datasets shows that the algorithm can effectively detect power line foreign objects and has application value for power line inspection tasks.