In order to realize removal function model building with high precision in computer controlled optical polishing process, a kind of new algorithm is proposed in this paper, which was based on edge detection technology. The algorithm preprocesses the interferometric metrology result and inverts it into digital image data at first. And then the Canny edge detecting operator and Hough transform are used to detect and extract the image edge. At last the boundary of the removal function was acquired. The center of removal function boundary was allocated quickly in this algorithm, which was assisted with the particularly center position. It reduced a lot of blind calculation for the target center of circle accumulating counts. At last, the experiment result indicated the algorithm in this paper could be used in removal function modeling and dwell time optimization solving.
1INTRODUCTIONSComputer controlled optical surfacing (CCOS) technology, which was proposed by Itek company (in American) in 1970s, is a kind of small tool polishing method [1][2] . This technology points out that target material removal quantity ( ) , F x y equal to the two-dimensional convolution of dwell time matrix ( ) , T x y and removal function ( ), R x y [3][4][5] . At present, people usually use theoretical modeling or experiment modeling method to obtain the removal function ( ) , R x y . Compared with theoretical modeling method,experiment modeling method is a kind of more accuracy approach to obtain removal function. The basic process of this method could be described like this: (1) Polishing a removal spot on the sample surface at first; (2) Measuring the full surface figure data by interferometry technology; (3) Extracting the removal area data based on the figure measuring result. In the process mentioned above we can obtain the removal function. So the key problem is how to obtain the edge or contour of the removed region in the interferometry dataset.Edge detection technology is one of important branches of image processing. A special object or a special part of object could be detected by this technology. Because of the rapid changing of gray value at the edge areas, the edge of object could be efficiently detected by the differential edge detection operators [6][7][8] . But, these operators have the disadvantage of improving a lot of noise data. Better than edge detection method, Hough transform could detect edges of the objects which have special geometric shape. The advantage of this algorithm is anti-interference ability to inhibit noise data efficiently. And the disadvantage is the computing speed is slow [9][10][11] .According to the removal function modeling problem in CCOS technology, removal function extraction algorithm is proposed in this paper based on image processing technology. At first, interferometric metrology result is converted to