概率引导的随机采样一致性算法  

Probability Guided Random Sample Consensus

作　　者：刘坤[1] 葛俊锋[1] 罗予频[1] 杨士元[1] 

机构地区：[1]清华大学自动化系清华信息科学与技术国家实验室(筹),北京100084

出　　处：《计算机辅助设计与图形学学报》2009年第5期657-662,共6页Journal of Computer-Aided Design & Computer Graphics

摘　　要：为了提高随机采样一致性算法的计算效率,提出一种概率引导的随机采样一致性算法.根据采样模型在原始数据上的检验结果调整每个样本点的采样概率,使得正确样本和正确模型被采样的概率得到提高.在首次获得正确模型之后,样本采样与模型更新构成了一个正反馈环节,经过若干次迭代后,正确样本被采样的概率远超过错误样本被采样的概率.理论分析和实验数据表明,该算法收敛需要的迭代次数较少,有效地提高了随机采样一致性算法的效率.In this paper a probability guided Random Sample Consensus algorithm is proposed in order to improve the computational efficiency of the traditional Random Sample Consensus. The weight of each sample is dynamically adjusted during the sampling process by measuring the errors of the estimated model on original data set, which enhances the probability of sampling a correct subset. The iteration process of sampling and model updating forms a positive feedback after a good subset is sampled. The inlier samples will be sampled with a much larger probahility than outlier samples after several iterations. The theoretical analysis and experimental results show that our proposed algorithm can obtain a good model estimation with less sampling times compared with the traditional Random Sample Consensus algorithm.

关 键 词：随机采样一致性 鲁棒性 模型估计 概率 基础矩阵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]