重合度约束的近景影像直线匹配算法  

Straight Line matching Algorithm of Close-Range Images based on Overlap Constraint

作　　者：朱红[1] 宋伟东[1] 杜北[2] 赵丽科 

机构地区：[1]辽宁工程技术大学测绘与地理科学学院,辽宁阜新123000 [2]哈尔滨工业大学土木工程学院,黑龙江哈尔滨150000 [3]武汉大学遥感信息工程学院,湖北武汉430079

出　　处：《信号处理》2015年第8期912-917,共6页Journal of Signal Processing

基　　金：国家自然科学基金项目(41101452);辽宁工程技术大学研究生科研立项资助(5A2014027-01)

摘　　要：针对目前近景影像中由于直线断裂或遮挡等干扰因素导致匹配的效果不理想,本文提出一种基于直线重合度相似性的近景影像直线匹配算法。该算法首先利用SIFT算子匹配同名点,并采用RANSAC算法对匹配结果进行优化,然后利用同名点计算仿射变换矩阵。通过仿射变换将左影像直线映射到右影像上,计算待匹配直线组线段的重合度,并对匹配结果进行优化,最终确定同名直线。实验证明,该算法原理简单,能够有效的避免繁琐计算,同时具有较好的鲁棒性,可以处理不同类型的近景影像数据。Focus on the unsatisfactory matching results which caused by the line fragmentation, occlusion and interference factors. An improved line matching algorithm based on the similarity of straight line overlap degree for close-range images is proposed in this paper. Firstly, use the SIFT matching algorithm to extract the corresponding points; secondly, use the cor- responding points to calculate the affine transformation; thirdly, map the left image line to the right image through affine transformation, calculate the overlap of the matching lines, and optimize the matching results ; finally, determine if they are the same line. Experiment results show that the algorithm is simple and effectively avoids complicated calculations. Also it has better robustness which means the applicability for different types of close-range images data.

关 键 词：直线匹配 直线重合度 仿射变换 近景影像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]