自适应三角形约束的多角度影像多基元匹配方法

项目名称： 自适应三角形约束的多角度影像多基元匹配方法

项目编号： No.41201472

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： 张云生

作者单位： 中南大学

项目金额： 25万元

中文摘要： 倾斜摄影测量是新一代摄影测量研究的国际学术前沿。倾斜摄影系统通过同时从一个垂直角度和多个倾斜角度采集影像，能够获取地形地物顶面和侧面的高分辨率影像，给三维重建、正射影像纠正带来了更多的信息，但多角度影像的可靠自动处理也给传统摄影测量带来了挑战。针对倾斜摄影测量的共性核心问题- - 多角度影像可靠匹配，本项目在深入分析多角度影像的成像几何、尺度差异和高分辨率等特点，并系统研究三维目标特征与多角度二维影像特征之间映射关系的基础上，提出自适应三角形约束的多角度影像多基元匹配方法，旨在突破传统方法匹配特征单一、可靠性较差的瓶颈，主要内容包括：（1）高重复率与高精度线特征提取；（2）线特征匹配约束条件与线特征匹配相似性测度；（3）自适应三角形约束的多角度影像点-线特征联合匹配传播；（4）多角度影像密集匹配。该研究将为多角度影像的自动化智能化处理、高精度三维重建、正射影像纠正等提供新的途径。

中文关键词： 倾斜摄影测量；特征提取与定位；影像匹配；数字表面模型；几何成

英文摘要： Oblique photogrammetry is an international academic frontier of the new generation of photogrammetry. By simultaneously collecting images from a vertical angle and sevral tilt angles, oblique photography system can obtain high-resolution top and side images of terrain and scene. It brings more information for three-dimensional reconstruction and orthophoto correction, however, reliable and automatic processing the multi-angle images arouses a new challenge to existing photogrammetry system. Aiming at reliable multi-angle image matching for the common problem of oblique photogrammetry, this project analyzed the sensor geometry, scale difference and high resolution characters of the multi-angle images comprehensively, as well as the mapping relationship between 3D object features and multi-angle 2D images features, then proposed a multiple primitives image matching method for multi-angle images constrained by self-adaptive triangulation, which is designed to overcome the bottleneck of traditional image matching method with single feature and low reliability. The main research contents include: 1) Line features extraction method with high repeatability and precision; 2）Constraints and similarity measure for line features matching; 3）Integrated point-and-edge matching propagation by using the self-adaptive triangula

英文关键词： Oblique Photogrammetry；Feature Extraction and Location；Image matching；Digital Surface Model；Image Geometry Model