一种基于形状描述的模型配准方法

New appoach to model registration via shape descriptor

针对模型配准中需要预配准操作的问题,提出了一种基于几何形状描述的模型配准方法。首先利用一种基于距离场的几何形状描述方法从预配准模型中获取特征点并建立对应关系,即构造预配准模型间的特征点对;之后,以特征点对为基础计算模型配准所需的转换矩阵以实现对模型的一步精确配准。实例表明该方法在不需要对模型进行预配准操作的情况下可实现模型的精确配准。

For model registration, this paper proposed a new method based on shape descriptor. First, the method extracted feature points on the given models by using a distance-based shape descriptor. Next, it calculated similarities of these feature points on different models and used for determination of feature point pair. Finally, it calculated the transformation matrix based on feature point pairs, and used the matrix to register the given models. The experimental results show that the proposed method can make a high quality registration without needing any pre-processing.