边界聚类椭圆快速检测方法

Fast ellipse detection based on edge grouping

针对多个椭圆因重叠、遮挡、嵌套情况而存在检测效率与精度较低的问题,提出一种基于边界聚类的椭圆快速检测改进方法.该方法包括图像预处理、边界聚类、椭圆拟合和去伪过程.进行图像预处理,包括边界检测、细化边界和消除冗余等价边界列（EELs）,去除多数非椭圆边界像素.开展边界聚类,通过边界像素连接、线段列提取和圆弧聚类,得到一系列候选椭圆弧和椭圆弧对.采用直接最小二乘方法拟合椭圆,并作去伪处理.利用椭圆的形态信息,调整算法步骤、优化阈值,提高算法效率,并通过实验评估算法性能.结果表明,边界聚类方法可准确、快速检测不同形态的椭圆,改进后用时可缩短14%~76%.

An improved ellipse detection method based on edge grouping was proposed due to the low detection efficiency of multiple,overlapping,occluded and nested ellipses.The method consisted of image preprocessing,edge grouping,elliptical fitting and removal of false alarms.First,the majority of non-elliptic boundary pixels were removed in the process of image preprocessing through edge detection,edge thinning and redundant EELs（equivalent edge lists）elimination.Then,a series of candidate elliptic arcs and elliptic arc pairs were obtained in the edge following process through edge linking,line-segment list extraction,neighborhood and global merging of arcs.Ellipse fitting was performed by using direct least-squares method,and the false alarms were subsequently removed.Besides,combined with the existing forms of target ellipses,the algorithm was simplified and the threshold was also optimized.Finally,the proposed method was evaluated by using both synthetic and real-world images.Results show that the edge grouping method can detect ellipses accurately and fast in different existing forms,with 14%~76%time shortened after optimization.