Flaw detection of cylindrical surfaces in PU-packing by using machine vision technique

Abstract

PU-Packing is widely used in machinery to prevent fluid from leaking. Due to the high demand for productivity and quality, bare-eye-inspection approach becomes extremely inadequate. In view of that, we developed a machine-vision-based system for detecting flaws occurred in PU-packings. The system consists of three inspection stations. The first station uses area-scan cameras to check the top and bottom surfaces of a packing. The second station uses an area-scan camera to check the interior of a packing incorporated with an infrared LED backlight. The third station uses two line-scan cameras to simultaneously scan inner and outer cylindrical surfaces of a packing. After images have been scanned, we applied variance of radius inspection method, projection method, coordinate transformation techniques, and normalized grayscale absolute difference inspection method to locate seven frequently detected defects, including deformations, burrs, interstices, scratches, blisters, air bubbles, and inclusions. The performance of the proposed system was evaluated by the accuracy, false positive rate, false negative rate and inspection time. The overall accuracy is 90.1% indicating that the proposed inspection apparatuses and methodologies are appropriate.