Research on the influence of image motion blur on the effectiveness of machine vision-based metal scraps separation system

Machine vision technique is gaining popularity in metal scraps recycling process due to the increasing need for automation and the relatively low cost of industrial camera. Relative movement between camera and scraps within camera exposure time will inevitably cause image motion blur and influence the separation accuracy. In this study, a novel deblurring prior for camera-conveyor system was proposed and applied to a metal separation system. Key performance indicators, like trajectory stability, separation distance, and separation accuracy, were analyzed and compared between blurred-case separation and deblurred-case separation. The results indicated that by applying the proposed deblurring algorithm, the Peak Signal-to-Noise Ratio (PSNR) value of all the blurred images increased, the average PSNR value of all the 31 blurred images is 27.71 and that of the deblurred images is 29.51. Objects center coordinates shift shows a significant decrease after the deblurring process. A higher conveyor speed may increase the separation efficiency, but will cause the decrease of separation accuracy and separation purity. The scraps motion trajectory obtained from deblurred cases is much more stable, which implies a stable landing point of scraps, and by applying the proposed deblurring algorithm, the separation accuracy increases from 82.9 to 88.0%, and separation purity increases from 90.7 to 93.6%.