Research On Defect Detection Of Bearing Sleeye Based On Image Processing

Machine components are the basis of the survival and development of the equipment manufacturing industry. As an extremely vital machine component, a high-quality bearing is necessary. The bearing producers care a lot about the detection of bearing's producing qualities. Nowadays, corporations adopt artificial methods to defect bearings. The load of work is not only heavy and inefficiently, but also high to leak the defects, so it affects the quality of bearings' dispatch. In order to solve the long-term problem, corporations urgently need the automatic equipment, instead of the manual manipulation. In this dissertation, static images of bearing sleeve are the main research objects. This paper designs a compact and low-cost embedded system based on ARM and CMOS. It integrates the functions of image acquisition and image processing, so it can effectively determine whether the bearing sleeve's chamfer is complete. The work of the paper focuses on the following aspects:1. Extensively investigate and study the image using in defect detection, the application of image processing, comprehend their respective advantages of the actual defect detection.2. The paper designs a compact and low-cost embedded image acquisition system based on STM32F103ZET6+OV7670+AL422B. It expands external data memory. So it can obtain an image and save it in IS61LV25616.3. On the basis of hardware platform, through programming the camera driver, this paper can configure the internal registers of the camera. Debugging the camera so that it can get a QVGA and RGB565 format image.4. By the use of principle of reflection of light, the image is different if the bearing sleeve has chamfering. Through image processing, the paper counts the black pixels in the chamfer region to judge defect. To reduce the computational complexity of image processing, the paper extracts the chamfer region to research. It extracts a rectangular region and an annular region. And after series of experiments, the paper compares the robustness and accuracy of the two methods. In the last, it chooses to extract an annular region to identify the defect.