Abstract
With the rapid development of the automotive industry, terminals-as critical components of wiring harnesses-play a pivotal role in ensuring the reliability and stability of signal transmission. At present, terminal crimping quality inspection (TCQI) primarily relies on manual visual examination, which suffers from low efficiency, high labor intensity, and susceptibility to missed detections. To address these challenges, this study proposes an improved YOLOv5-based model, TCQI-YOLOv5, designed to achieve efficient and accurate automatic detection of terminal crimping quality. In the feature extraction module, the model integrates the C2f structure, FasterNet module, and Efficient Multi-scale Attention (EMA) attention mechanism, enhancing its capability to identify small targets and subtle defects. Moreover, the SIOU loss function is employed to replace the traditional IOU, thereby improving the localization accuracy of predicted bounding boxes. Experimental results demonstrate that TCQI-YOLOv5 significantly improves recognition ccuracy for difficult-to-detect defects such as shallow insulation crimps, achieving a mean average precision (mAP) of 98.3%, outperforming comparative models. Furthermore, the detection speed meets the requirements of real-time industrial applications, indicating strong potential for practical deployment.