基于Shape Context的字符识别算法介绍

本算法提出了一种新的计算形状相似度的方法，并且把这种方法应用到目标识别领域中。

这中相似度衡量的方法主要经过以下几个步骤：

(1)在两个形状上找到匹配的点对

(2)使用匹配的点对估计两个形状的形变程度。为了解决匹配点对的问题，本算法中引入了一个描述向量，即Shape Context，每一个点都有一个描述向量。Shape Context

以某个点为参考点，统计其他的点在他周围的分布特征，根据这个分布特征，就可以知道该点所处的全局特征。在两个形状上，如果是匹配点对，那么他们的Shape Context应该是相似的。

本文采用了Shape Context的这样一中相似度度量方式，来进行字符是识别，取得了良好的效果。-Shape Context-based character recognition algorithm introduced in this algorithm presents a new method of calculating the similarity of shape, and put this method to target recognition field. This is the main method of similarity measure through the following steps: (1) find the matching points on the two shapes (2) the degree of deformation using the matching points on two shapes estimate. In order to solve the problem of matching points, the algorithm is described in the introduction of a vector, i.e., Shape Context, each point has a descr iptive vector. Shape Context to a point as a reference point, the statistical distribution of other points around him, according to this distribution, we can know that the point in which the global feature. On both the shape and, if the matching point, their Shape Context should be similar. In this paper, in such a similarity metric Shape Context, and to identify the character, and achieved good results.