An Efficient and Lightweight Surface Defect Detection Method for Micro-Motor Commutators in Complex Industrial Scenarios Based on the CLS-YOLO Network

• DOI: 10.3390/electronics14030505
• Publication Date: 2025-01-27T14:42:23Z
• Authors: Qipeng Chen; Qiaoqiao Xiong; Haisong Huang; Saihong Tang
• Keywords: CLs upper limits

Existing surface defect detection methods for micro-motor commutators suffer from low accuracy, poor real-time performance, and high false missed rates small targets. To address these issues, this paper proposes a high-performance robust commutator model (CLS-YOLO), using YOLOv11-n as the baseline model. First, lightweight Cross-Scale Feature Fusion Module (CCFM) is introduced to integrate features different scales, enhancing model’s adaptability scale variations ability detect objects. This approach reduces parameters improves speed without compromising accuracy. Second, Large Separable Kernel Attention (LSKA) module incorporated into head strengthen feature understanding capture, reducing interference complex patterns on significantly improving various target types. Finally, issues related center point location, aspect ratio, angle, sample imbalance in bounding boxes, SIoU Loss replaces CIoU original network, overcoming limitations of loss function overall performance. Model performance was evaluated compared dataset, with additional experiments designed verify effectiveness feasibility. Experimental results show that, YOLOv11-n, CLS-YOLO achieves 2.08% improvement mAP@0.5. demonstrates that can accurately large targets while maintaining accuracy tiny defects. Additionally, outperforms most YOLO-series models, including YOLOv8-n YOLOv10-n. The parameter count only 1.860 million, lower than increase 8.34%, making it suitable deployment resource-limited terminal devices industrial scenarios.