Deep learning-based automated optical inspection system for crimp connections

Within the trend of electrification and autonomous driving, the significance of high-quality crimp connectors is increasing as they establish the electrical connection for the energy and information flow in the automotive system. Whereas the manufacturing of crimp connectors is highly automated, the final quality assessment mainly comprises manual optical inspection tasks that are human labor-intensive and time-consuming. Addressing this gap, a computer vision system to automate the final inspection of crimp connectors is proposed and implemented. In this paper, the image processing chain and the deep learning-based model to reason over image data of crimp connectors with regard to different defect classes are outlined. The effectiveness of this system using a dataset collected in the laboratory environment is demonstrated.