摘要
提出一种在数字图像中估计椭圆模型参数的鲁棒方法。该方法采用自下而上的思路,结合RANSAC(Random sample consensus)算法,先将图像中的样本分割成具有类内结构相似性的子类群,再依据类间拟合相似性将子类合并,最后在完成聚合的类中估计出模型参数。该方法的优势在于无约束性,不需要先验条件,可以在模型的数量、尺度等信息未知的情况下进行参数估计,并有效抑制离群数据影响。实验结果表明,该方法估计精度较高,鲁棒性能良好。
Based on the ellipse model,a novel robust algorithm is proposed to estimate the model parameters from the digital image.Firstly,all the sample points are divided into subclasses containing similar fitting property by combining with the random sample consensus(RANSAC) algorithm.Then,subclasses are mutually combined according to different fitting properties.Finally,the model parameters are estimated from the compounded subclasses.The proposed method shows its advantage with nonrestraint.It can estimate parameters when the scale and the quantity of models are unknown and can restrain the noise.Experimental results show that the method has accurate accuracy and good robustness.
出处
《数据采集与处理》
CSCD
北大核心
2010年第3期407-412,共6页
Journal of Data Acquisition and Processing
基金
国家"八六三"高技术研究发展计划(2007AA04Z227)资助项目
模式识别国家重点实验室开放课题基金(2006-3)资助项目
关键词
椭圆模型
鲁棒估计
RANSAC算法
样本拟合
ellipse model
robust estimation
random sample consensus(RANSAC) algorithm
sample fitting