Defect Detection Method for Power Communication Fibre Optic Cables Based on Improved YOLOv8

In response to the problems of limited accuracy, inability to effectively identify subtle defects, and insufficient ability to handle different input resolutions and aspect ratios in defect detection of power overhead communication optical cables, this paper proposes a defect detection method for power communication optical cables based on improved YOLOv8. This method introduces a Global Connectivity Part (GCP) module into the Neck network to enhance intra-layer feature interaction and enrich feature details. By replacing it with a dynamic detection head, it automatically adapts to the input shape to generate anchor boxes, significantly improving the recognition accuracy and flexibility of small targets such as line damage, loose fittings, and other defects. The experimental results show that compared with the original YOLOv8 and other mainstream algorithms, this method not only performs well in reducing false positives on the dataset provided by the State Grid Jilin Electric Power Co., Ltd. Tonghua Power Supply Company, but also effectively detects the vast majority of defects, reducing the possibility of missed detections. Its accuracy, recall rate, and mAP@50 have all been significantly improved, demonstrating good practical application potential.