基于五点算法估计基础矩阵的研究  

Research of Fundamental Matrix Based on 5-Point Algorithm

作　　者：周春艳[1] 范学栋[1] 邹峥嵘[1] 

机构地区：[1]中南大学信息科学与工程学院,湖南长沙410083

出　　处：《计算机技术与发展》2011年第11期8-10,共3页Computer Technology and Development

基　　金：国家自然科学基金(6107187)

摘　　要：基础矩阵是三维重建、运动估计、图像定标、匹配的基础,是解决计算机视觉和图像处理领域的重要课题。文中利用点到极线的距离最小原理对五点算法进行改进:在初始模型估计中,使用五点算法,降低算法的抽样次数和抽样时间,然后使用点到极线的距离最小原理对五点算法的多项式的伪解进行剔除,利用正确解求得本质矩阵,再根据本质矩阵与基础矩阵的归一化关系得到基础矩阵。实验证明此算法提高了估计基础矩阵的准确性,排除错误解,提高了五点算法的正确率。Fundamental matrix is the basis of three - dimensional reconstruction and motion estimation,camera calibration and matching , is an important task in computer vision. It uses the Polar distance minimum principle to 5-point algorithm for improving：in the initial estimation, using the S-point algorithm, reducing the number of algorithm of sampling and sample time, and then point to a line of minimum distance principle removed to 5 points algorithms for polynomial false solution, use the correct solution to the nature of matrix, and then by the basic nature of matrix and matrix normalization of relations between fundamental matrix get fundamental matrix. Experiment proved the algorithm improved the accuracy of estimation basic matrix and removed misunderstanding solution,improved the accuracy rate of the 5-point algorithm.

关 键 词：基础矩阵 本质矩阵 五点算法 点到极线的距离最小原理 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]