边缘视差连续性约束的航空影像特征线匹配算法

鉴于同一边缘线上像素的视差变化具有连续性的特点，本文提出边缘视差连续性约束的航空影像特征线匹配算法。该算法首先对左影像进行特征线提取，将Freeman链码跟踪得到的每条边缘特征线作为一个整体，利用相互连接直线段对其进行拟合。将拟合直线端点作为该特征线上的一组匹配基元，通过对其进行匹配实现该特征线的整体匹配。初始匹配过程中，综合利用核线约束、单应矩阵约束和灰度相似性约束对拟合直线的端点进行匹配。然后利用视差连续性约束对每条特征线上点的匹配结果进行检核及余点同名点构建，取得较好匹配结果。论文选取UCX和DMC三组数字航空影像进行特征线匹配实验，验证本文算法的有效性。 

Feature line matching algorithm for aerial image based on continuity constraint of edge parallax

Given the parallax changes of pixels in the same edge have the characteristics of continuity, a feature line matching algorithm for aerial image based on continuity constraint of edge parallax is proposed in this paper. Firstly, it extracts feature lines from left image, and considers each edge feature line obtained by Freeman Chain Code Tracking as a union, which is fitted by the connected line segments. And then it takes the endpoints of fitting line as a set of matching primitives, and makes the whole matching of feature line with these primitives. In initial matching process, it matches the endpoints of fitting straight line combining with epipolar constraint, homography constraint and gray similarity constraint. And then checks the matching results of points in each feature line and constructs corresponding points to those who have not been successful in matching using the parallax continuity constraint, and obtains good matching results. This paper selects three groups of digital aerial image form the UCX and DMC for feature line matching experiments, and the results verify the validity of algorithm in this paper.