对极几何估计的鲁棒性新算法  

A New Robust Algorithm for Epipolar Geometry Estimation

作　　者：陈泽志[1] 吴成柯[1] 刘勇[1] 

机构地区：[1]西安电子科技大学综合业务网理论及关键技术国家重点实验室,西安710071

出　　处：《计算机学报》2000年第6期634-639,共6页Chinese Journal of Computers

基　　金：国家自然科学基金!( 69972 0 3 9);中法先进研究计划项目

摘　　要：以 8参数模型的 F矩阵为基础 ,深入研究了 F阵的稳定性求解中的若干问题 .将求解对极点的问题归结为求解二元六次非线性方程的最优解问题 ,并将所求最优解作为进一步计算的最佳候选对极点 .最后以余差最小准则获得 F阵的全局最优解 .实验结果表明 ,该方法不但使对极点的稳定性有较大提高 ,而且也能高精度地估计出Based on the eight parameter model of F matrix, some problems on robust estimating fundamental matrix are studied. Unlike traditional method, such as, 6 point synthetic algorithm, the Least Square algorithm and the Least Median of Squares method, an optimization problem of nonlinear equation for solving epipoles is generalized in this paper. The optimal solutions of epipoles are the first rank candidate epipoles for further calculation, and the F matrix with minimal value of residual errors are regarded as the global optimal solution. Experimental results show the presented algorithm performs in terms of robustness, and an accurate fundamental matrix and two stable epipoles could be found.

关 键 词：鲁棒性 三维重建 图像分析 对极几何估计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]