Transmission Line Insulator Defect Detection Method Based on YOLO-MLSL Model

To address the challenges of insufficient small target recognition, difficulty in edge information extraction, and high computational overhead in insulator defect detection, this paper proposes a lightweight detection method based on the YOLO-MLSL model for transmission line insulator defect detection. First, a C2f-LRFCA module is introduced, effectively enhancing feature interaction through a long-range convolutional attention mechanism, thereby improving the perception of fine-grained defects. Second, an MEUM multi-scale feature enhancement module is designed to achieve more efficient contextual information fusion during upsampling, improving the detection performance for multi-scale targets. Third, the ShapeIoU loss function is employed to improve the bounding box regression accuracy in complex backgrounds, and LAMP pruning technology significantly reduces the model’s computational and storage overhead. Experimental results show that the improved algorithm achieves an [email protected] of 85.4%, a 4.1% improvement compared to the original YOLOv8n, while maintaining a low parameter count and computational complexity, demonstrating both high accuracy and efficiency. This research provides a valuable reference for the design and application of lightweight target detection models in the intelligent inspection of power equipment.