用于识别一类缺陷字符的过筛连通模板匹配法

Sifting-connecting Templet Matching Algorithm Applied to Disfigurement Character Recognition

针对一类发生切割的缺陷字符,传统的模板匹配法难以对其进行正确识别,基于连通域的改进算法也存在一定的误识率。文章提""出了一种基于过筛连通的模板匹配法较好地解决了这一问题。算法对基于连通域的模板匹配法进行了进一步改进,其基本思路是对缺陷字符采用先过筛,再连通,最后匹配识别的办法,最大程度地保留了笔道上的黑像素,屏蔽了笔道外的干扰黑像素,从而实现字符的识别。该算法构思新颖、计算简洁,大量实验表明其识别率也是可以保证的。

In the light of a kind of incised disfigurement character, traditional templet matching algorithm can't exactly recognize it, and the mended algorithm based on connecting-field also presents some error-recognition ratio. This paper brings forward a templet matching based on sifting-connecting, which preferably solves this problem. The algorithm has more improvement , whose basic thought is firstly to sift disfigurement character, then to connect, and at last match to recognize. This algorithm, which furthest reserves the black pixels in the character-path, and shields the interferential black pixels out of the character-path, realizes the character recognition. It designs novelly, and calculates succinctly. Many experiments indicate that the recognition ratio is credible.