中文摘要：

提出一种基于图象序列对应概率累积的基础矩阵求解算法.该算法基于序列对应概率累积求最大熵点的对应概率累积分布图,采用RANSAC思想和加权最小二乘法拟合极线,通过统计分类器求解极点并消除出格极线,并利用所求的对应极线和极点,采用一种参量化方法求解F矩阵.该算法工程实现容易,对输入的源图象没有苛刻的要求,可以基于随机拍摄的图象序列方便地求解F矩阵.

英文摘要：

An algorithm for solving fundamental matrix based on accumulated sequence correspondence probability is proposed,which has no rigorous requirement on inputting images.It can solve F-matrix based on random image sequences,so its realization is very easy.The idea is that,the accumulated probability chart corresponding to maximum entropy point is plotted.Epipolar line is fitted by RANSAC and weighted least square method,and the outlier epipolar lines are removed according to the epipole solved by statistical classifier.And the correspondent epipole and epipolar line is gained in the same way.