Abstract: To avoid the disturbance of flaws and improve the accuracy in measurement of surface cleanliness-level of precise optical elements,an image detection system based on machine vision method was proposed,in which a set of movable clamping fixtures for optical elements of various sizes and a three-dimensional electronic control platform were designed.In image processing,Canny edge detection operator was used to extract image edges,and the convex hull method was employed to obtain enclosed areas for the detected object.Then,a test vector was composed by the characteristic parameters extracted from the geometric space,gray-level space and spectrum domain of the enclosed area,and analyzed by relevance vector machine(RVM) method to recognize particle and non-particle contaminants on surface.Based on the results,the cleanliness-level grade of the precision optical elements can be obtained.