基于多特征关联的复杂地形高分辨率遥感图像匹配技术研究

项目名称： 基于多特征关联的复杂地形高分辨率遥感图像匹配技术研究

项目编号： No.41201349

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 地理学

项目作者： 杨健

作者单位： 中国科学院遥感与数字地球研究所

项目金额： 25万元

中文摘要： 针对大范围区域遥感应用中，复杂地形条件下的多传感器遥感影像数据配准拼接要求，考虑不同成像角度、不同尺度以及地物高程偏移影响因素，以高空间分辨率的卫星影像和无人机航空影像为研究对象，结合面向对象处理方法进行地物对象的多重特征提取与分析，提出适用于起伏条件的局部不变量点特征和线特征的分布约束，研究基于多尺度几何分析的点、线特征提取技术，建立基于地物的多重特征松弛匹配策略，并结合局部匹配与样条形变模型改进重叠区域的拼接线自动选择技术，支持区域地物精细修正，解决多种特征关联的、面向多角度、局部拼接的特征匹配关键技术问题，实现对多种传感器数据源影像的亚像元自动配准拼接。

中文关键词： 影像配准；多角度；特征提取与定位；多尺度几何分析；薄板样条

英文摘要： For the use requirements of multi-sensorremote sensing image registeing in large areas remote sensing applications such as land cover monitoring，study the Practical Algorithm under complex terrain conditions，considering the different imaging point of view, different scales, as well as surface features elevation offset influencing factors. Appling the high spatial resolution satelliteimages and UAV aerial images,we extraction and analysis the multiple features on covering objects combined with object-oriented approach. In the research, we approach a method of the distribution of constraints that apply to the complex terrain conditions of local invariant point features and line features, and a method of multi-scale geometric point, line and feature extraction techniques based on MGA. Thus to establish a modal of refined SOR algorithm based on covering objects. Finally, automaticly select the stitching line of the overlap region with the matching relation of local objects and the Spline correction modal, and refine the local correction model to improve accuracy.So to resolve the problems of feature registration on a variety of characteristics extracted, multi-angle and local correction.And achieve the object of sub-pixel, automatic registration on the variety of sensor data source image.

英文关键词： image registration；multi-angle；feature extraction；MGA；Thin-plate spline