图象配准是计算机视觉中目标识别的一种基本方法.其目的是在待识别图象中寻找与模型图象的最佳匹配.该文以传统的Umeyama点集相关度量为基础,结合Procrustes正规化方法,通过引入加权矩阵,以得到新的相关度量函数,进而提出了一种图象特征点集匹配的新方法,解决了传统方法要求点集维数相同的缺点.经过迭代运算,对存在几何失真,且维数不同的两点集可得到精确配准.文中给出的点集配准结果说明,当两点集维数相同时,该方法不仅与传统的点集相关法一样,均可达到精确配准,而且该方法对维数不同的点集也可实现精确配准

Image registration is a fundamental object recognition method in computer vision. It aims to find a best match of an object image in an image to be processed. In this paper, we concentrate on image registration from image feature point-sets. A new method is proposed which is based on the conventional correlation measure of two point-sets which was introduced by Umeyama. The traditional Procrustes analysis method is used to normalize the point-sets. The novelty of the proposed method is by introducing a weight matrix into Umeyama' s correlation measure the limitation of the traditional method, which requires the dimensions of both point-sets to be the same, is released. The proposed method can register two point-sets with geometrical distortion and different dimensions. Point-sets registration results are given in the paper. When the dimensions of both point-sets are the same, both of the proposed method and the traditional method work well. But when the dimensions are different, only the proposed method can register point-sets precisely.