中文摘要：

针对多视影像由于遮挡导致的匹配不确定性问题,该文提出一种特征点和物方地面元相结合的多视影像密集匹配方法。该方法利用规则格网划分的空间平面作为基础,通过特征点匹配获取同名像点和高程赋值的格网平面,为后续匹配提供初始的DSM。在此基础上,利用铅垂线轨迹法对平面上规则分布的平面元进行匹配,加密特征点匹配结果。通过对某区域UCX影像的匹配实验表明,该文提出的算法能高效、准确地得到匹配结果,特别对较大影像计算快速,适用于实时处理。

英文摘要：

Aiming at the uncertainty problem of multi-view image matching due to the occlusion, this paper presented a multi-view image matching method combining feature points and ground elements in ob- ject space. The method used space plane that are divided by regular grid as the foundation. First of all, the corresponding image points and the planes that had the same elevation were got through matching the fea- ture points to provide the initial DSM for the subsequent matching. Secondly, on the basis of the previous step, the plane elements that were of rules were matched by combining with the vertical line locus method to infill feature point matching results. Finally, making the matching experiment by aerial images to verify the validity of the proposed method. Through a regional UCX image matching experiments, it showed that the proposed algorithm could get matching results efficiently and accurately, the computation was fast es- pecially for larger image, and suitable for real-time processing.