针对远程水表抄表系统中的水表数字识别问题,提出一种有效的水表数字图像二次识别算法.在识别之前,先对水表图像进行前期处理以获得单个字符图像.基于模板匹配的思想,该算法首先将待识别字符与模板只进行有效信息区比较,得到一次识别匹配系数;然后进行基于数字字符特征的匹配,得到二次识别匹配系数,最后取2次匹配系数之和最大者为最佳匹配.对于半字识别,在变化字长模板匹配的基础上,提出匹配权重的概念,并结合上下半字连续这个先验知识,将待识别半字图片分别与各个模板匹配,取匹配系数最大者为最佳匹配.实验结果表明,所提方法简单有效且识别率很高,获得了满意的效果.
According to the digital identification problem of water meter in the remote meter reading system,an effective tw o-times recognition algorithm about meter digital images is proposed.Singlecharacter images are obtained through pretreatment on the meter image before recognition.Based on the idea of template matching,the algorithm first compares the character images to be recognized w ith the templates only on their effective information area to obtain the first recognition matching coefficient.Secondly,the second recognition matching coefficient is conducted on the basis of the digital character feature matching.Finally,the maximum sum of tw o matching coefficients is taken as the best matching.For half-w ord recognition,based on w ord-changing height template matching,the concept of matching w eight is proposed.Combining the priori conditions that the upper and low er half-w ords are continuous,the half-w ord image to be recognized is compared w ith every template and the largest matching coefficient is taken as the best matching.The experimental results show that the proposed method achieves satisfactory results and it is simple and effective w ith high recognition rate.