Abstract: This paper presents a method of 3D reconstruction for outdoor scene from a single catadioptric omnidirectional image. The method can reconstruct the 3D model of all the scenes in the catadioptric omnidirectional image and support walkthrough. To implement the reconstruction, this paper firstly classifies the catadioptric omnidirectional image to horizontal ground, vertical building surface and vertical background surface and gets the structure of the scenes in the catadioptric omnidirectional image. Using the above classification, the paper can implement the 3D reconstruction based on the catadioptric model. The result shows that the method has the advantages of easy source image capturing, big view of field and completely automatic processing, furthermore, the method is also available for non-planar scene.

Key words: catadioptric omnidirectional image, remote sensing image, 3D reconstruction, images matching, walkthrough

摘要： 提出了一种针对室外场景的单幅折反射全向图三维重构方法, 能够自动重构出全向图中360视野内景物的三维模型, 并实现自由漫游。基于全向图与遥感图匹配把全向图分为水平地面、垂直建筑物立面和垂直背景景物面三类区域, 得到全向图场景的基本结构; 在此基础上利用折反射光路投射模型计算出全向图中每个像素点的三维几何位置, 从而实现了折反射全向图的重构。实验证明该方法具有采集简单、视野大、处理过程全自动化、能够重构非平面场景等特点。

关键词: 折反射全向图, 遥感图像, 三维重构, 图像配准, 漫游