基于霍夫变换椭圆检测的两种改进算法

针对一维参数空间霍夫变换椭圆检测算法不能检测长轴端点缺损的椭圆这一缺陷,提出了两种改进算法.两种改进算法都是先确定椭圆中心,采用分步计算参数的方法,以降低霍夫变换参数空间的维数,从而降低算法整体的时间复杂度.改进算法1可以看成是一维霍夫变换椭圆检测的一个补充算法,它通过改为最后对长轴进行一维累积,来解决长轴端点缺损的问题.改进算法2则是在椭圆中心确定后,利用椭圆的对称性,计算出其倾斜角,再进一步确定其他参数.实验表明,两种改进算法都能在较复杂的图中较快较准确地检测出残缺椭圆.

Two Improved Algorithms for Ellipse Detection Based on Hough Transform

Two improved algorithms are proposed to solve such a problem that the ellipse detection algorithm based on Hough transform in one-dimensional parametric space can not detect ellipses when the endpoints of the major axes are not available. Both the two improved algorithms firstly determine the center of the ellipse, and adopts the method of calculating the parameters step by step in order to reduce the dimension of the parameter space of the Hough transform, so as to reduce the overall time complexity of the algorithm. The improved algorithm 1 can be regarded as a supplementary algorithm for the ellipse detection of one-dimension Hough transform, and is used to solve the problem of major axis endpoints defect by adopting the one-dimension accumulation of the major axis. The improved algorithm 2 calculates the inclination angle of an ellipse by using the symmetry of the ellipse, after finding the center of the ellipse, and further other parameters of the ellipse are determined. The experimental results show that the two improved algorithms can all detect the incomplete ellipse rapidly and accurately in complex environment.