Abstract: Aiming at the problems of low efficiency and high error rate in artificial classification of bearing house,the industrial camera calibration,image acquisition of bearing house, hybrid denoising,feature recognition and target classification were studied.At first,the images of bearing houses were collected under industrial lighting by image acquisition system.Median filter was designed to filter the noise of images. Feature filtering was used for further hybrid denoising of images after binaryzation processing. Then, bearing houses contours were fitted and the minimum outsourcing rectangle were calculated,these images were put straight by affine transformation.Finally, they were categorized according to the specific characteristics, and the capture experiment was carried out by EPSON robot. The results indicate that the overall recognition accuracy of the method is up to 99.967%, which is better than traditional manual detection. The system can meet the requirements of industrial production.
Key words: machine vision; OpenCV; bearing house; recognition system