Substation Equipment Defect Detection Algorithm Based on Improved YOLOv8

In the defect detection of substation equipment, there are problems such as complex scenarios, high false detection rate and missed detection rate of small-size defects. In view of the above requirements, this paper proposes a defect detection algorithm for substation equipment based on the improved YOLOv8, which exceeds the accuracy of the mainstream object detection model on the substation defect dataset. Firstly, EfficientFormerV2 serves as the backbone network for the enhanced model, which enhances the ability of the model to extract local and global information. Secondly, A dedicated small object detection layer has been engineered to facilitate the identification of defects in tiny targets measuring 4×4 units. Finally,The LSKA attention mechanism is employed to bolster the model's capacity for multi-scale feature extraction. Experimental results indicate that this approach, when compared to YOLOv8n, the accuracy of the proposed algorithm is increased by 2.48%, and The number of parameters and the count of floating point operations have experienced only a slight improvement. The enhanced algorithm significantly boosts the efficiency of identifying defects in substation equipment, fulfilling the requirements for such defect detection.