Research On Surface Defect Detection Technique Of Micro Bearing Based On Computer Vision

The bearing is a very important part of the mechanical industry. As it is applied to many fields, the processing precision and quality of the bearing concern service function and life-span of mechanical product. Therefore, the detection of bearing's producing qualities has been the question which the bearing producers care a lot all the time. There are a few bearings whose surfaces have defects or nicks, etc. They become useless products. Therefore, the workers have to reject out the useless products or the substandard products from the products before they dispatch them out from the factories. So they must inspect the bearings one by one carefully. Nowadays corporations adopt artificial methods to complete the work of detecting the defected bearings. The load of word is not merely heavy and inefficiently, abut also high to leak the defects, so it affects the quality of the dispatch. In order to solve the long existing problem, corporations urgently need the automatic detection instrument, instead of the manual manipulation. For the computer visual system can obtain a large amount of information and is apt to integrate designing information and process information. The computer visual system that tests work condition surveillance and the worksites surveillance already has applied to industry and the commerce field broadly.Ningbo is a base of our country's bearing. It has more than 1,000 bearing enterprises. The Ci Xi Import and Export Corporation exports bearings account for 10% of the whole nation bearing exportation. Bearing has already becomes the important foreign exchange earning product in Ningbo city. In recent years, every enterprise has accomplished the automatic manufacture of bearing through the technological revolution. However, it is still the manpower visual measurement to detect to the surface defect and to reject them. This method is not only inefficiently but also low reliable. Therefore, the bearing producers also need this automatic detection instrument urgently.In the foreign countries, the computer visual system had been widely used in the automatic visual detection and automatic identification of the industrial parts. Nowadays, computer visual inspection has been used as a test of the appearances of the products and the surface defect, such as metal surface visual inspection, defect detecting test of printed circuit board, automatic welded seam identification and the belt conveyer visual inspection etc..Domestic visual detecting studies gain more attention just in the 90s, but it has achieved some important development by leaps and bounds in the recent years. Some universities and scientific research institutions have invented some computer visual detection system. In the decades, the visual detection technologies have achieved great progresses, and there are more and more achievements in the theoretical research and in the applied research. Although we have gained some achievements, we are still in the starting stage and fall behind the other countries comparatively.The paper combines the project of industrial science and technology of Zhejiang province (2007C31014) and the project of industrial science and technology of Ningbo city(2005B100014)- the study of detecting the surface defect of micro bearing based on CCD image recognition. It adopts computer skills, sensor technology, communication technology, digital image analysis technology to accomplish the functions of delivering, handling, controlling, etc. It has a very important significance in detecting the suface defect of micro bearing. The main detection system in the thesis is composed of CCD, photoelectric sensor, frame grabber, computer, display instrument and manipulated system etc.The thesis discusses the structures of the visual detection's hardware and its selected principles, analyzes the functions of light sources, and design the lighting system. It also illustrates the function evaluation index of the visual detection technology. It points out that hardware is the prerequisite and basis of the visual detection. This thesis mainly based on the technological requirements of the parts and the present condition to select the proper parts to satisfy the aim of visual detection.Digital image's conventional treatment is the basic theory of computer vision inspection. The paper analyses noise source and property in digital image, and suggests some general methods to suppress noise. It also takes more space to illustrate the edge detection and image threshold segmentation. In the aspect of the edge detection, it mainly discusses the common edge detection operator, Canny edge detection operator. In the aspect of image thresholding, it mainly discusses the peak-valley method,minimum erro split and maximum variance threshold method. It also compares the advantages and the disadvantages of the detection time and the detection effect of the edge detection method through the experiment, so that the paper selects Canny operator to detect edges.In order to segment the image, it has to detect the center of the micro bearing. It gives a brief introduction to the general Hough transform. At the same time, it suggests to use the three-point method to detect the center of the circle coordinate, this method promotes the speed of detecting the center of a circle greatly. Comparing this method to the general Hough method, it guarantees the precision of the detection, reduces the processes and the time of arithmetic. The three-point position circle method can also detect the disintegrated circle and the center of an arc. It broadens the range of the detection.The paper treats the inside lane and outside lane as a whole in detecting the surface of the bearing. Detecting surface defect of micro bearing based on difference image method not only will reduce the processes of the arithmetic but also can achieve a satisfied effect. Detecting the surface of micro bearing with characters uses of image matching.The detecting area is broadly located by the center of micro bearing, and exactly located by sectering characters. Then, match the objectibe area and detect it to judge the bearing of qualified or not. Through the detecting to the 500 bearings'experiments, visual detection's precision to the flawed bearings can reach up to the 99.5%. It achieves the desired result.