Inspection of Automobile Components Using Mask R-CNN and Minimal Setup Vision

Detecting surface defects in anodized automobile components on a serial production line is a formidable task that demands precision, given theseparts' critical role in ensuring vehicle safety and performance. This research proposes a ground-breaking system that tackles these challenges by minimizing hardware and leveraging software processing, rendering it automated, versatile, affordable, and precise. The proposed method adopts the Mask R-CNN algorithm for defect detection, enabling pixel-level instance segmentation and bounding the defects with boxes. Moreover, a multi-label classifier is deployed to distinguish between defective and non-defective components and predict the type of defect. The system's design also accounts for the computational time needed to ensure the system is suitable for high-speed conveyor belts. By accurately detecting defects and preventing defective components from causing catastrophic consequences, the proposed approach has the potential to revolutionize defect detection in serial production lines, making the process faster, more reliable, and safer