首页 | 本学科首页   官方微博 | 高级检索  
     

对极几何估计的鲁棒性新算法
引用本文:陈泽志,吴成柯,刘勇. 对极几何估计的鲁棒性新算法[J]. 计算机学报, 2000, 23(6): 634-639
作者姓名:陈泽志  吴成柯  刘勇
作者单位: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
基金项目: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,深入研究的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上步计算的最佳修造对极点,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,也能高精度地估计祟F阵。

关 键 词:鲁棒性 三维重建 图像分析 对极几何估计

A New Robust Algorithm for Epipolar Geometry Estimation
CHEN Ze-Zhi,WU Cheng-Ke,LIU Yong. A New Robust Algorithm for Epipolar Geometry Estimation[J]. Chinese Journal of Computers, 2000, 23(6): 634-639
Authors:CHEN Ze-Zhi  WU Cheng-Ke  LIU Yong
Abstract: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.
Keywords:fundamental matrix   robustness   3D reconstruction   residual error   singular value decomposition(SVD)
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号