基于改进最小平方中值算法的基本矩阵估计算法

Method for Estimating Fundamental Matrix based on Improved LMedS Algorithm

针对基本矩阵的估计问题,为了消除误匹配点对和高斯噪声的影响,本文提出了一种估计基本矩阵的改进算法——基于遗传算法的最小平方中值算法,该算法是在最小平方中值算法中引入遗传算法,克服了传统鲁棒算法中对内点集的数据等同处理问题,利用遗传算法的全局优化特性,通过遗传算法对最小平方中值算法所得到的内点集合进行筛选,利用选取的8个匹配点对来估计基本矩阵。通过实例仿真表明,该算法在误匹配点对和高斯噪声存在的情况下具有更好的鲁棒性和精确性。

Aimed at the evaluation problem of fundamental matrix,and for eliminating the impacts of mismatching point and Gaussian noise,the paper proposed a new robust method （LMedS＋GA）based on genetic algorithm,which input GA in LMedS and can overcome the problems of handling of traditional robust algorithm to inner point data.Genetic algorithm has good global optimization,first of all,the method of LMedS tests and excludes the initial matched points,then,defined each matched point for a gene and a chromosome consist of eight genes.A minimal subset of the fundamental matrix is a chromo-some.Through simulation,the method has better robust and accuracy under the condition of existing mismatching point and Gaussian noise.