An Automated Vision-Based Inspection System for Bearing Gland Covers

This paper presents an automatic vision-based system for bearing gland cover quality control. The system employs the method of gradually refined scheme to locate regions of interests. Several types of defects are detected from their corresponding regions by utilizing image segmentation, curve fitting, feature validation, and other image processing methods. Although each technology is not strange to us, how to integrate them into an entire inspection system efficiently and effectively is a huge challenge. In addition, some useful visual features such as maximum of orientation difference and maximum rectangular feature are proposed to validate candidate defects. Field tests demonstrate that the proposed system gains an excellent performance.