Abstract:  Detection of artificial circular target center is a key in 3D measurements. In order to get accurate calibration and measurement in industrial fields, a robust algorithm for circular target center detection is proposed. Images with circular target are first preprocessed, and Canny edges extracted. The canny edges
are then refined with an improved Snake algorithm, and sub-pixel edges are found using a Zernike operator. Accurate centers are obtained through robust ellipse fitting. Compared with traditional methods, the proposed algorithm can give accurate center detect results even if the target is damaged or soiled. Simulation results and real data show that the center detection error is less than 0.02 pixel in ideal cases, and less than 0.03 pixel when the target is damaged.

Key words: circular target, sub-pixel, ellipse fitting, data registration