中文摘要：

为提高求解精度,提出一种基于改进的随机抽样一致性（RANSAC）算法的基础矩阵求解方法。采用加权策略,将局内点占全部匹配点的比例作为权重函数的自变量;利用本质矩阵和基础矩阵的关系,鉴于本质矩阵两个非零奇异值应该相等这个特性,利用加权因子和本质矩阵的奇异值构造目标函数,这两点改进意味着目标函数中有两个约束条件的限制;利用matlab遗传算法工具箱来求解目标函数的最小值,可以得到准确的基础矩阵。模版图像实验和场景图像实验验证了该算法的有效性。

英文摘要：

To improve the precision,a solving method of fundamental matrix based on improved random sample consensus algorithm（RANSAC）was proposed.weighted strategy was used and the inliers ratio was served as the independent variables of the weighting function.Given the relationship between fundamental matrix and the essential matrix and the nature that two non-zero singular values of the essential matrix supposed to be equal,the method was proposed based on the essential matrix characteristics,in addition,the weighted strategy was used.The two improvements means the objective function was limited,ensuring the overall accuracy of calculating parameters,then a cost function was constructed,lastly matlab genetic algorithm toolbox was used to solve the function of global minimum,in which way accurate fundamental matrix was gotten.The effectiveness of the algorithm proposed was verified by the template image experiment and scene image experiment.