同名点累计分布约束的航空影像线特征匹配

为了克服断裂、变形等因素对线特征匹配的影响,本文提出了一种同名点累计分布约束的航空影像线特征匹配算法。该算法在Canny边缘提取的基础上对两张影像中的待匹配线特征建立索引,首先利用单应性矩阵结合核线约束精简候选线特征数量,并通过以点代线的方式对线特征上的所有点进行匹配;然后建立累计分布矩阵统计同名点的分布情况以确定同名线特征;最终整合双向匹配结果,实现线特征的多对多匹配。选取典型的航空影像进行线特征匹配实验,结果表明:本文提出的匹配算法能进行曲线匹配,并可实现断裂线特征的多对多匹配,鲁棒性强。与已有匹配算法进行比较,该算法在尺度、亮度、旋转等变化条件下仍可取得较为满意的匹配结果,且在匹配效率上具有一定的优势。

Line Matching for Aerial Image Based on Cumulative Distribution Constraint of the Corresponding Points

In order to overcome the influence of fracture, deformation and other factors on line matching, a novel method is proposed for line matching based on cumulative distribution constraint of the corresponding points. The method first indexes the lines extracted by canny algorithm, then the candidate lines are determined by the homo-graph matrix and the epipolar constraints. Matching all the points on the lines, a cumulative distribution matrix is built to record the distribution of the corresponding points, which determines the corresponding lines. Finally, the many-to-many matching are achieved by integrating the bidirectional line matching results. Extensive experiments on aerial images demonstrate thatbetter result has been achieved by the proposed methods in curve matching and many-to-many matching. Compared with the existing matching method, the method hasaccurate line matchingunder the changes of scale, brightness, rotation and so on, and it also has advantages in the matching efficiency.