基于拓扑图及三维特征的三维物体识别方法

Recognition of 3-D Objects Based on Topology and 3-D Attributes

本文提出一种基于拓扑图及三维特征的三维物体识别方法。该方法在图像预处理阶段首先检测物体可视表面的闭合边界,依据闭合边界对物体进行分割,并用一个景物特征图对景物进行表达(SAGR)。在建模阶段,采用多视图表面模型方法进行建模,依据表面分割,对多视图模型用一个模型特征图进行表达(MAGR)。在物体识别阶段,将景体与模型的匹配分成三部分进行,即扫描部件、图的匹配部件及匹配检验部件。匹配首先是基于SAGR与MAGR图的拓扑性质的匹配,然后是基于三维特征的匹配。根据匹配结果,或对未知物体自动建模,或给出识别结果。

A new 3D objects recognition approach based on topology and 3-D attributes is presented. In the preprocessing stage, the enclosed edges of visible object surfaces are detected first and a scene attribute graph to represent the scen(?) (SAGR) is used. In modeling stage, a multiple-view surface model is applied to build a model represented by a model attribute graph (MAGR) according to surface segmentation. In object recognition stage, the approach divides the matching between scene and model into three parts, i.e. the Scanner, the Graph matcher and the Matching Verifier. The matching is based on the topology at first, then on the 3-D attributes between SAGR and MAGR. According to the matching the approach either automatically builds model for unknown scene or gives the recognition results. The effectiveness of the approach presented in this paper has been shown in the experiments.