Abstract: This paper describes an algorithm framework for building recognition in urban area image sequence. Three image sequences and DSM from multi-view match are used as sources of information. A homologue feature matching, based on geometry constraint, is applied for automatically generating 3D point and line. Features are grouped based on perceptual organization criteria. And an optimized texture extraction and mapping strategy are discussed.

Key words: 3D reconstruction, Image sequence, Match, Image processing

摘要： 提出了一种从序列图像建立城市三维模型的算法框架。该方法使用三条序列影像和匹配生成的数字表面模型，结合几何约束的多视特征匹配，从图像获取三维点、线信息。同时讨论了基于感知编组的几何模型生成算法，以及纹理优选与自动映射。实验结果证明该方法是有效的。

关键词: 三维重建, 序列影像, 匹配, 图像处理

CLC Number: 

• TP391.41