Research On Key Technology Of Visual Inspection Of Ceramic Ball Surface Defects

Ceramic balls are a kind of basic spherical parts with mirror-like characteristics.They are often used as bearing rolling elements in important fields such as aero engine and precision electric spindle.The development of an advanced manufacturing industry calls for much stricter requirements on precision grade and yield for ceramic balls.Therefore,the surface quality of ceramic balls must be rigorously inspected at factory.The artificial detection method,which is widely used at present,cannot meet the needs of actual production,but there is still no mature automated inspection equipment at home and abroad.Thus,it is of great significance to research and develop efficient,stable and accurate detection system for surface defects of ceramic balls.Based on the machine vision technology,two kinds of detection methods based on image gray difference and fringe reflection are proposed and verified experimentally.The key technology of design for light source,distortion correction of reflected fringes,image processing algorithms and so on have been deeply researched.In this paper,research is done to analyze the optical properties of ceramic balls and improve the existing dome diffuse light source.In order to improve the contrast of defects in images,a kind of light source with the use of uniform scattering light is proposed.An actual light source device is designed according to the size of ceramic ball.And the images of ceramic balls are acquired in experiments.According to the single peak distribution of gray value of ceramic ball images,an algorithm used for selecting threshold-segmentation parameter is proposed and the effectiveness of the proposed algorithm is verified.Some surface defects are small or not significantly different from normal surface in gray-value.And these defects cannot be effectively extracted by grayscale threshold segmentation method.Aiming at this practical problem,a high precision detection method based on fringe reflection is proposed in this paper.In this paper,a calibration calculation model for the fringe pattern on reflected screen is proposed to eliminate the distortion caused by the spherical contour of ceramic ball,so that the fringe patterns in images of ceramic balls are evenly distributed.According to the images with evenly distributed fringe patterns,two kinds of algorithms based on edge extraction and eigenvalue calculation are proposed for extracting defects.In this paper,silicon nitride ceramic balls with 6.35 mm diameter are selected as experimental object.Reflected screen are designed and experimental system are built to acquire images.For different types of defects in images,the proposed algorithm can effectively extract the defective areas and provide specific parameters of defective areas.Thus,the effectiveness and feasibility of proposed method for inspecting surface defects of ceramic balls have been verified.