Research On Surface Defect Detection Of Micro Bearings Based On The Computer Vision

Some defects, such as little pits, nicks and scratching and so on, may be caused on its surfaces during the bearings are produced and assembled. Because the size, depth and position of the above defects are often random, the most common way to pick out the defected bearings is to check every bearing one by one just by the workers' eyes. It not only results in the heavier workloads and lower efficiency but also seriously affects the supplying quality of bearings for the inevitably higher rate to lose detection. The computer visual technology has been widely used in the automatic detection systems in the advanced countries, and it is obvious that untouched detecting the micro bearings surface based on CCD image recognition is the useful way to solve the problem mentioned above.The paper relies on the project, "The Study of Detecting The Surface Defect of Micro Bearing Based on CCD Image Recognition", supported by both the Key Technology R&D Program in Zhejiang Province(No.2007C31014) and in Ningbo municipal(No. 2005B100014). The designed detection system is composed of CCD, photoelectric sensors, frame grabber, computer, display instruments and manipulated system etc. Using synthetically the technologies of computer, sensor, communication, digital image analysis to obtain the functions such as information obtaining, delivering, handling and controlling, the system can detect the micro bearings' surface defect at the real-time.The general theories about Computer Vision Inspection(CVI), the filter methods to suppress noise, the technologies about the edge detection and image threshold segmentation, all kinds of edge detection operators, are discussed in the paper. And on the basis of experiments and comparing the detection time and the detection effect synthetically, the Canny Edges Detection Operator is selected to use in the project. The paper put forward a method, Three-Point Method(TPM), to detection the center coordinate of a circle. Comparing to the general Hough method, TPM can ensure the precision of the detection, speed the circle center detecting greatly, simplify the process and reduce the time of arithmetic.The paper treats the inside lane and outside lane of the bearing as a whole during detecting the surface. The way to detect the above area based on the Difference Image Method not only reduce the processes of the arithmetic but also can achieve a satisfied effect. The Image Matching Algorithm is used to detect the bearing's sealed cover area which has some characters on. Firstly the detected area is broadly located by the center of micro bearing, and then it is exactly located based on the affection on the grey value statistics in the sector area by the characters, and at last the aimed area is detected and the quality of the bearing is judged.A device has been designed to carry micro bearings and pick out the defected ones automatically too, which is made up by PLC and pneumatic controlling system.The experiments show that the result made by the above detection system is satisfying. So the work on the research and development presented in the whole paper is meaningful to improve the level of the Computer Visual Inspection technique, and it is also valuable to the practical application of the CVI technology.