基于Hough变换的椭圆检测算法

为了实现光电仪器对椭圆形目标的准确识别与跟踪，基于Hough变换提出了一种新的椭圆检测算法。该算法随机采样2点，再利用椭圆极和极弦的性质来搜索第3点并筛除大量无效采样；然后，以这3点为中心作正方形窗口，用窗口内的所有点来拟合椭圆方程。在验证候选椭圆时，提出了一种新方法来判断边缘点是否在椭圆上，并且给出了确定真实椭圆的自适应阈值。实验显示，该算法的平均长度误差为0．5pixel，平均角度误差为0．6°，平均耗时为79ms，表明该算法精度高，速度快，检测性能较好。

Ellipse detection algorithm based on Hough transform

In order to ensure that photoelectric instruments can identify and track elliptical objects accurately, a new algorithm based on Hough transform is proposed. The new algorithm randomly samples two points, and then searches the third point using the characters of ellipse's pole and pole chord, and eliminates lots of invalid samples. In the following, it uses the three points as the centers to make three square windows, and then all the points in the windows are used to fit the ellipse. When a candidate ellipse is validated, a new method is proposed to judge if edge points are on the ellipse, and an adaptive threshold is supplied to confirm real ellipses. The experiment indicates that the algorithm's average length error is 0.5 pixel, average angle error is 0.6°, and the average time needed is 79 ms. In conclusion, the algorithm has high precision and high speed, and shows a good capability of detecting ellipses.