车辆牌照字符识别

介绍了一个牌照字符识别系统的字符串图像提取部分和字符识别部分 .前者采用基于纹理的分割阈值选取方法 ,即先对牌照图像进行纹理分析 ,然后对取得的直方图采用模式识别中的最大最小准则决定分割阈值 ,其效果优于传统的直方图方法 .后者采用多个简单识别器融合方法进行字符识别 .这一方法根据不同类型识别器的特性 ,采用串、并联混合的方案将它们组合在一起 .与单一识别器相比 ,不仅大幅提高了识别率而且缩短了识别耗时 .这一系统对牌照汉字、数字和字母的识别率均大于 96%,总的牌照识别率为 82 %.

Number-Plate Character Recognition

This paper mainly described the string picking- up module and character recognition module of a number plate recognition system.The former acquired the basic information through texture analysis, then determined the segmentation threshold using Mini- Max criterion.The result of this method is obvi- ously superior to the pre- existed method,which was directly based on grey- level histogram of an image. The latter module combined several simple recognizers together to utilize the peculiarity of each one.As a result,its recognition rate is higher and recognition time is shorter than thatof each single recognizer em- bedded in it.The individual character recognition rate of this system is over96%and the whole license plate recognition rate is 82 %.