A Circle Detection Algorithm Based on Ellipse Removal

In optical CCD detection, due to distortion, the circle will appear elliptical shape after projection onto a two-dimensional plane through perspective. In order to solve this problem, a circle detection algorithm based on ellipse de-falsification was proposed. The image was preprocessed by filtering, the axial ratio of the distorted circle was set, and the ten points randomly selected on the contour of the image were used to determine whether the circle was within the reasonable distortion range. Quadratic interpolation method was used to detect the sub-pixel edge of the contour point set, and based on the principle of Random Sampling Consensus (RANSAC), the outliers outside the threshold range were removed to achieve the effect of false elimination. Finally, the distorted circle was fitted by the least square method. Experimental results show that the detection error of this method is about 0.3%.