Defect Detection Method for 10kV Column Equipment Based on Superposition Segmentation and Improved YOLOv8

To address the challenges of small recognition targets and significant background interference in the detection of defects in 10kV column equipment, a novel defect detection method based on superposition segmentation and an improved YOLOv8 model is proposed. To tackle the issue of small target detection, the proposed method introduces a superposition segmentation approach, which divides the detection image into four equal parts and incorporates the golden section ratio. This strategy effectively enlarges the recognition targets and enhances detection accuracy. Furthermore, to mitigate the limitations of the original YOLOv8 model, which is prone to background interference and inaccurate target localization, the proposed method integrates the GAM attention mechanism and replaces the original loss function. These modifications increase the weighting ratio of the target, thereby improving detection performance. The method was evaluated on a self-constructed dataset, named the CCDD (Column Equipment Detection Dataset). Experimental results demonstrate that the proposed method achieves an average detection accuracy of 86.7% and a recall rate of 83.2% for defect detection. Compared to the original YOLOv8 model, the accuracy is improved by 9.2%, and the [email protected] is increased by 15.5%. This approach provides a promising solution for defect detection in 10kV column equipment.