Research On Defect Detection Of Valve Sleeve Inner Hole Based On Machine Vision

With the rapid development of computer technology, machine vision and image processing technology have also achieved a raid development. More and more defect detection of the product can be achieved by machine, which not only liberates the workers from the heavy manual labor, but also improves the precision and speed of detection. This thesis detects the inner hole image information of the diverter valve sleeve based on the digital image processing technology.Diverter valve sleeve is the core component of the automobile steering system, its quality will directly affect the steering system of vehicle. This thesis describes an image stitching method based on two-dimension optical scanning according to the structural characteristic of the diverter valve pocket. Firstly, the information of the inner hole image should be abstracted to acquire inner hole image. This thesis designs a mechanical structure for collecting the information of inner hole image.The image should be preprocessed after the image collection, during which makes the image feature expressed obviously through image enhancement. Next, the de-noising of image is accomplished to remove the noise produced in the process of image collecting. Then, segmental process is done to make the image edge more obvious. Finally, image distortion correcting is done to facilitate the following processing. After the preprocessing, image stitching is done and the characteristic of image is extracted. Then, image feature is matched according to the extracted characteristic, after which image fusion is done through overlap region. Finally, a complete image of good quality was chosen to be a template within the complete inner hole image of valve pocket formed through stitching. Then the differential image is obtained through computing the difference of the templet image and the image which will be detected. Next, after researching the image segmentation, the binary differential image is got. Finally, the number of pixels in the defect area within the binary image was analyzed. If the number of pixels exceeds a certain amount, it will be believed to there exist defect in the image to be detected.