一种直线段匹配的新方法

A New Algorism of Matching Line Segments

直线特征在人造目标和环境中普遍存在,直线段的匹配是三维重建等应用领域的首要解决问题。一般单直线的匹配方法难以解决被部分遮挡直线的匹配问题,提出了一种基于极线约束和RANSAC算法的匹配方法。对直线段上的点运用基于极线约束的匹配方法在另一幅图像中进行点的同名匹配,采用RANSAC算法建立直线模型,剔除野值,得到匹配直线段。本方法能有效地解决图像中被部分遮挡直线段的匹配问题,仿真实验和实际图像实验证实了本算法的正确性和高精度。

Line segments are ordinary in man-made targets and matching line segments is the first step in a lot of fields, such as 3D reconstruction. Because of occlusion, the question of matching individual line segments is difficult to figure out with the traditional methods, and a novel approach based on epipolar-line constraint and the RANSAC algorithm is presented. Firstly, the points on line segments between image pairs are matched by epipolar-line constraint; Secondly, the RANSAC algorithm is used to estimate a straight line model fit to a set of 2D points, so the line segments of occlusion is matched availably. The result of the synthetic and real images is accurate and reliable.