IDST-YOLOv8: Insulator Small Target Defect Automatic Detection Model

The normal operation of insulators is an important link to ensure the safe and reliable operation of power system. In order to promote the automatic detection of insulator defects, a small target defect detection model IDST-YOLOv8 for insulator flashover and breakage is proposed. Firstly, the CSPC module is designed to improve the detection of low resolution images and small targets. Secondly, the AM-SPPF module with spatial pyramid structure is designed to make the model pay more attention to the global background information. Then, the CTA attention mechanism is added to effectively improve the detection performance of small target defects on insulators. Finally, a double-scale detection head is designed to make the model pay more attention to small and medium-sized semantic information. The experimental results show that the parameter number of IDST-YOLOv8 network detection model is 2.8M, which is 6.7% lower than that of the benchmark model; the value of insulator flashover and damaged mAP@0.5 is 95.8% and 93.8%, respectively, which is 3.2% and 6.8% higher than that of the benchmark model. Insulator flashover and damaged mAP@0.5:0.95 are 51.2% and 58%, respectively, which are respectively increased by 4.7% and 6.6% compared with the benchmark model. IDST-YOLOv8 promotes the deployment of intelligent detection algorithms on UAV equipment to facilitate the real-time detection of insulator defects.