实时字符识别在视频读表系统中的应用

为了保证字符识别的准确性，正确的字符分割是关键。由于光照不均匀等外界客观条件的影响，通常是对所采集到的数字图像先进行静态的各种预处理，然后才能进行识别从而得到正确的结果。这里详细地阐述了一种对采集到的非均匀亮度的数字水表图像上数字及符号的实时字符识别算法，并介绍了基于高斯滤波、投影算法、边缘检测等算法的图像预处理方法。通过现场实验，该算法的字符识别正确率达到100％，且实时性高。

Real-Time Character Recognition Method in the Video Reading System

In order to ensure the veracity of the character recognition, it ispivotal that the characters are partitioned accurately. Because of the uneven illumination affection of the external environment conditions, the acquired digital images are usually preprocessed statically firstly. Then the correct results can be gained by the algorithms of recognition. This paper expatiates detailedly the real-time algorithms of character recognition to the numbers and symbols of the acquired uneven brightness images of the digital water meters. And it also introduces a method of digital image preprocessing based on Gauss filtering, projection segmentation, edge detection and so on. Through lots of real experiments, it is verified that the correct distinguish ratio for characters has well achieved 100% and real-time of the algorithms is better.