期刊论文

Li, C.(lichang@mail.ccnu.edu.cn)

中文

武汉大学学报(信息科学版)

1671-8860

2010

35

12

1461-1465

10.13203/j.whugis2010.12.027

国家教育部科学技术重点资助项目（108098） 国家自然科学基金资助项目（40671078 40771088）

本校为第一且通讯机构

提出了一种新的基于直线自动分组、核线约束、灭点方向引导和影像金字塔模型的直线匹配算法。该算法利用核线约束缩小搜索范围但不完全依赖于核线,利用灭点方向引导匹配但不完全依赖于灭点,并且符合空间物体整体分布不连续但局部连续的客观规律,再加上大范围的搜索,不仅克服了一般直线匹配算法中核线约束退化或基于单应性映射失效的缺陷,而且能匹配出直线段端点。同时,还提出了基于随机抽样一致性的单片剔除误匹配直线的算法。实验验证了此算法的实时性、准确性和实用性。

Street elevation 3D reconstruction needs taking automation into account, therefore we present a novel straight line stereo matching algorithm based on straight line auto-grouping, epipolar line constraint guiding, vanishing point direction guiding and image pyramid model. The algorithm based on independent observation + local integral constraint +large area search, and uses epipolar line to reduce searched area but does not absolutely depend on epipolar line, which utilizes vanishing points' directions to guide line matching but does not absolutely depend on vanishing points' directions. Furth...