Insulator Defect Detection Algorithm Based on EGF-YOLO

With the expansion and aging of power transmission lines, the risk of insulator defects such as damage and cracks has increased. Traditional manual inspection methods are inefficient and prone to false negatives and false positives. To tackle the challenges of low precision and the occurrence of missed and erroneous detections in algorithms designed for identifying insulator defects, an improved version of the YOLOv8n algorithm, named EGF-YOLO, has been introduced. First, the ECA attention module is introduced into the C2f module of YOLOv8n, creating the C2f_ECA module. This module enhances the model's ability to focus on defect features, thereby improving the detection performance of defect characteristics.Second, the Conv module is replaced with the GSConv module, which reduces model complexity while enhancing the ability to capture detailed defect features. Finally, Focaler-IoU is introduced to replace the CIoU loss function of YOLOv8n to improve the regression accuracy of the model in detecting bounding boxes. Experimental results show that, on the self-made insulator defect dataset, the improved algorithm achieves a mAP of 89.1% and a inference speed of 140 FPS, which is an increase of 3.1% and 12 FPS compared to the baseline YOLOv8n, respectively. In comparison with other algorithms, the improved YOLOv8n algorithm outperforms others in terms of higher detection accuracy and faster detection speed, validating the effectiveness and performance advantages of the proposed algorithm.