Abstract: The license plate binarization technique has been a key issue in plate recognition system.A novel and effective method is proposed in this paper based on the idea of classification.The binarization problem is considered as a classification problem from the idea of statistical discriminate analysis.First of all，sampling from the contracted region，and then classify the pixels.In order to improve the accuracy of binarization，an iterative classification technique is proposed.In addition，a new index system is proposed in order to evaluate the effect of plate binarization，which can be used to evaluate the effect of binarization of plate license.The experimental results show that the binarization algorithm is simple and effective.

摘要： 对车牌区域进行二值化一直是车牌识别系统的一个关键问题。针对车牌区域的特征，提出了一种基于分类思想的二值化方法。该算法从统计判别分析的思想出发，将二值化问题看成是一个分类问题。首先对区域进行收缩取样，然后进行分类。为了提高二值化精度，其中还使用了迭代分类技术。另外为了评价车牌二值化效果，从车牌二值化应用角度出发提出了粘连度、字符断裂度、噪声颗粒数、运行消耗时间的指标体系，用来评价车牌二值化的效果。有了这套指标体系，就可以方便地对各种车牌二值化技术进行评价。实验结果表明，该二值化算法简单有效。