Aiming at solving the problem that the traditional circle detection algorithm is characterized by low detection speed and so is unfit for multicircle detection, a circle detection method based on global searching is proposed. By combining the accumulation of evidence with weighted average, the pseudocenter of a circle produced in the process of accumulation of evidence is classified and analyze, then 3 pseudocenters of a circle kind by kind are rejected and the other circle parameters are computed. The experimental results show that the proposed algorithm is of high efficiency and is not sensitive to the defect of local information. And the detection time will not increase linearly with the numbers of circle and the detection result is better than the traditional random circle detection (RCD) algorithm obviously.