Research on Defect Detection Algorithm of Substation Equipment based on Improved YOLOV8

Aiming at the problem that the existing target detection algorithms have insufficient recognition performance for small target defects such as metal corrosion and insulator damage in substation equipment, a defect detection algorithm based on improved YOLOv8 is proposed. Firstly, a frequency domain feature fusion module (FDFFM) is designed to enhance the global perception of texture features through dual-path feature extraction of Fourier transform and spatial convolution; secondly, a composite coordinate attention module (CCAM) is embedded in the backbone network to improve the positioning accuracy of small targets in complex backgrounds through long-distance dependency modeling; finally, the bounding box regression loss function is optimized to alleviate the problem of overfitting in training. Experimental results show that the mean average precision (mAP) of the proposed algorithm on the self-built substation equipment defect dataset reaches 68.8%, which is 2.5% higher than the baseline model YOLOv8. In particular, in the detection of metal corrosion and insulator damage defects, the AP values are increased by 2.8% and 1.2% respectively.