一种POS数据辅助多视角倾斜航空影像匹配方法  

A POS Supported Matching Method for Multi-View Oblique Aerial Images

作　　者：赵明衍[1] 戴晨光[1,2] 狄亚南 仇多兵 张伟政[3] 

机构地区：[1]信息工程大学,河南郑州450001 [2]武汉大学测绘遥感信息工程国家重点实验室,湖北武汉430079 [3]国家电网河南省电力公司郑州供电公司,河南郑州450001

出　　处：《测绘科学技术学报》2016年第4期431-435,共5页Journal of Geomatics Science and Technology

基　　金：国家海洋公益基金项目(201305002);国家自然科学基金项目(41571336);中央高校基本科研业务费专项(3132015002;3132014034;3132016003);中央高校基本科研业务费创新团队(3132014302)

摘　　要：提出一种利用POS辅助多视角倾斜影像匹配的算法。首先,利用POS数据对倾斜影像进行近似核线纠正;然后,用SIFTGPU算法对纠正影像进行特征匹配,根据匹配结果计算出两张影像的水平和垂直方向视差进而求得其近似重叠区域;将重叠区域进行分块特征匹配,采用比值提纯法、视差约束和RANSAC算法等约束条件剔除误匹配;最后,将匹配结果通过POS数据反算回到原始影像上。试验结果表明,将POS数据应用到多视角倾斜影像匹配中,可快速获得比常规影像匹配方法数量更多、分布更均匀的匹配点。A POS supported matching method for multi-view aerial oblique images is put forward. First of all, the POS data is used to rectify the oblique images by epipoliar geometry constraint. Then SIFTGPU algorithm is used for feature matching on the rectified images. According to the matching results, the parallax of horizontal and verti-cal direction is computed to get their approximate overlapping area. Thirdly, the block feature matching is appiled to the overlap areas and the constraint conditions such as nearest neighbor distance ratio, parallax constraints, RANSAC algorithm are applied to eliminate false matches. In the course of time, match results are counted through POS data back to the original images. The result of experiment shows that matching points can be quickly obtained more in quantity and more uniform in distribution compared to the conventional image matching method when the POS data is applied to the multi-view aerial oblique image matching.

关 键 词：倾斜摄影测量 影像匹配 外方位元素 特征匹配 近似核线影像 

分 类 号：P228[天文地球—大地测量学与测量工程]