Research on Insulator Defect Detection Based on Synthetic Fog and Improved YOLOv5

To improve the accuracy and robustness of insulator defect detection in challenging environments, an enhanced YOLOv5-based detection algorithm is proposed. The model incorporates an improved Convolutional Block Attention Module alongside the SE attention mechanism and refines the BottleneckCSP structure to enhance feature extraction and channel weight distribution. The modified CBAM is integrated into the YOLOv5 backbone, enhancing the model's capability to capture essential features. Furthermore, an optimized image fogging algorithm is introduced, leveraging multi-threaded parallelization to accelerate processing. A synthetic dataset containing fogged images is generated to strengthen the model's performance in difficult environments. Results from experiments highlight that the proposed algorithm substantially boosts detection accuracy and bounding box regression stability, particularly in complex backgrounds and severe weather conditions. The model achieves an overall mAP50 of 99.4%, with 99.3% for normal insulators and 99.5% for defective ones, reflecting respective improvements of 0.3%, 0.5%, and 0.3% over the baseline YOLOv5 model. This approach enhances insulator defect detection precision and offers a dependable technical solution for industrial applications such as power transmission line inspections.