The safe operation of electrical equipment in substations is of paramount importance. Building upon the research foundation of the YOLOv5 object detection model, this paper introduces two innovative improvements. Firstly, a lightweight network model is adopted to replace the traditional backbone feature extraction network, thereby reducing model parameters and accelerating model training. Additionally, the Channel and Spatial Attention Mechanism (CBAM) is incorporated to enhance the feature information extraction capability, resulting in a significant improvement in detection accuracy. Through these enhancements, a dual enhancement in detection accuracy and recognition speed is achieved. This method enables real-time defect detection of substation equipment, replacing the manual inspection by on-site personnel and reducing their workload. Not only does it lower labor costs, but it also enhances operational efficiency. Experimental results demonstrate that the improved algorithm exhibits faster detection speed and higher accuracy compared to traditional algorithms.
Liu et al. (Wed,) studied this question.