中文摘要：

针对近景影像中直线因遮挡、提取算法等带来的匹配问题，提出了一种基于局部仿射不变及核线约束的直线特征匹配方法。该方法在对立体近景影像进行同名点匹配及直线特征提取的基础上，首先利用直线特征与邻域内以同名点为端点的虚拟直线段的相交仿射不变性进行候选同名直线的搜索，再借助核线约束确定目标直线与候选同名直线的重叠部分，然后通过直线特征的角度及支撑域灰度的综合相似性测度进行匹配，得到正确的同名直线。利用该方法对不同几何变换下的近景影像进行实验，结果表明，本方法具有较好的可靠性和准确性。

英文摘要：

To solve the line matching problem caused by obstruction and extractton algont,ml Iu range images, a novel algorithm of matching straight line based on local affine invariant and epipolar constraint is presented in this paper. On the basis of straight line extraction and corresponding point matching on dose-range image pairs, the linear feature＇s candidate corresponding lines are firstly searched based on the intersection affine invariant character of the lines including the extracted and virtual straight ones, while endpoints of the virtual are corresponding points near the extracted lines. Then, the overlapping parts of the target lines and the matching candidates are calculated based on epipolar constraint. The right corresponding lines are finally acquired with the help of the comprehensive similarity measure derived from angle and neighborhood grey level of the straight line. Using the approach proposed in this paper, experiments were carried out for the close-range images under different geometric transformation. The results show that this method has a good reliability and accuracy.