Advancing Automobile Manufacturing Inspections: YOLO-Based Missing Part Detection

In manufacturing, the precise assembly of critical components such as screws, bolts, caps, and connectors is vital for ensuring product integrity and safety. Traditional manual inspection methods for detecting missing parts are inefficient and error-prone, posing significant challenges to maintaining high-quality standards. Existing studies often focus on single-class components, lacking the ability to comprehensively address the complexity and variety of multi-class components in many manufacturing processes. To tackle these issues, we propose an automated approach to detecting missing parts in various manufacturing processes by leveraging advanced computer vision techniques, specifically the YOLOv8 model. Our system continuously monitors the assembly process using high-resolution imaging and real-time data processing to identify discrepancies. We introduced a new multi-class dataset specifically designed for missing parts detection in automobile fuel tank assemblies. This dataset includes 1323 annotated images covering various components, enhancing the model’s generalization ability across different scenarios. Our experiments demonstrated that the YOLOv8 model significantly outperforms the Faster R-CNN model across various performance indicators. The YOLOv8 model achieved a precision of 97.8%, recall of 98.2%, mAP@50 of 98.4%, showcasing its superior accuracy and efficiency. This research highlights the potential of automation to enhance quality control and reduce production costs across diverse manufacturing applications, proving the effectiveness of our proposed methodology in real-world manufacturing environments. The Missing Parts (MP) dataset is available at Roboflow