多种字符混合图像的自动识别

Recognition of Multi-character Hybrid Images

以货运列车自重识别系统为例 ,提出一种适应不同光照环境的图像增强方法 ,利用模糊集理论从多行信息中提取自重行 ,正确率接近 99%。在字符识别中 ,采用 3种神经网络分类器分别识别汉字、数字和英文字母 ,并对易混数字采用两级分类器的结构 ,获得了较高的识别率。

A new adaptive image contrast stretching method, word line segmentation and object region selection using fuzzy principle are proposed. In the character recognition, this paper presents three artificial neutral network classifiers trained by improved back propagation algorithm by which Chinese word, ten numeric digits(0～9) and the letter't' is recognized respectively. The two grade classifier is proposed to recognize the easy confusing numeris digits and the great recognition ratio has been achieved.