Drainage Pipeline Defect Detection Method Based on an Improved YOLO11

The prediction of defects in urban drainage pipelines plays a crucial role in improving urban water safety and ensuring residents' daily lives. Existing YOLO-based models for detecting defects in drainage pipelines suffer from low accuracy in complex backgrounds, often leading to false detections and missed detections. This paper proposes an improved YOLO11 algorithm. By optimizing the bottleneck architecture within the C3k2 module and introducing a lightweight Coordinate Attention module, the model effectively enhances the ability to extract defect features. In addition, this paper replaces the original classification loss function with the Adaptive Threshold Focal Loss (ATFL), enabling the model to focus more on defect features, reduce false detections caused by background interference, and thereby improve detection accuracy. The results show that, compared with the base YOLO11 algorithm, the improved algorithm proposed in this paper not only improves the detection precision but also enhances the accuracy and recall.