基于单应矩阵的直线匹配  

Matching Straight Lines on the Basis of Homography

作　　者：娄安颖[1] 宋伟东[1] 刘薇 

机构地区：[1]辽宁工程技术大学测绘与地理科学学院,阜新123000 [2]城乡建设管理局,阜新123000

出　　处：《遥感信息》2011年第3期9-13,共5页Remote Sensing Information

基　　金：基于异源遥感数据的三维重建模型研究(40771159);面向复杂建筑物部件的地面激光扫描点云与近景影像混合三维建模方法研究(40901222)

摘　　要：提出了在单应矩阵原理约束下进行直线匹配的算法。首先利用单应矩阵将从不同影像上提取出的直线集变换到同一影像坐标系下;然后结合直线自身的几何特性及其特性支持区域的灰度特性,初步确定同名直线;最后依据左右一致性准则,剔除错误匹配。当引入直线特性支持区域的灰度测度时,只考虑几何变形和灰度畸变较小的一侧,提高了匹配的可靠性。实验表明,该算法简单、易行,并有效解决了部分被遮挡直线以及断裂直线的匹配问题。A novel algorithm of matching straight lines based on homography is presented in this paper.In the first instance,it transforms the sets of straight line segments extracted form different images to the same image coordinate system.After that,combining the own geometric properties of straight lines and gray properties of their property support region,it ascertains the matched straight lines preliminarily.At the end,the mismatches were excluded with cross-correlation criterion.When it imports gray measure of property support region of lines,it only gets whose geometry and gray distortions are small,thus this method improves the reliability of matching.With the experiments it is concluded that the algorithm proposed in this article is simple and easy,and it clears up matching problems of the lines which are partially blocked out and the lines which are broken.

关 键 词：单应矩阵 直线匹配 几何特性 灰度特性 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]