Defect Detection Algorithm for Electrical Substation Equipment Based on Improved YOLOv10n

Electrical substations are fundamental infrastructure components for urban operations and civilian life. Many substations located in remote areas face challenges in monitoring equipment defects because robots or drones cannot effectively access these substations for direct inspection. Addressing the shortcomings of current algorithms in small target detection, complex background processing, and the balance between real-time and accuracy, this paper proposes several key improvements to the YOLOv10n framework to address these limitations. First, we design a novel attention mechanism module that enhances model focus on target objects under adverse weather conditions. Second, we modify the SPPF module by incorporating average pooling operations to compensate for feature information extraction deficiencies caused by maximum pooling operations. Third, we develop an improved C2f-OD module to strengthen the backbone network’s feature extraction capabilities. Finally, we implement Focal EIoU as the loss function to accelerate convergence and minimize losses. Our method can effectively identify and detect different substation equipment defects. Experimental results demonstrate a detection accuracy of 96.64% across six equipment categories. These results confirm the effectiveness of our proposed method, which helps reduce the cost of equipment maintenance and replacement.