基于权系数标识符矩阵的车牌字符识别方法

License plate character recognition based on weight coefficient identifier matrix

为了提高对车牌字符的准确识别能力,提出了一种基于权系数标识符矩阵的模板匹配车牌字符识别方法。具体方法是在进行字符识别前为每一个车牌字符制定一个标准化的模板,再将每一个模板字符的像素依据像素区域、像素边缘区域和非像素及非像素边缘区域等标记成不同的区域,并依此为基准生成一个模板矩阵。根据车牌字符闭合区域个数及字符二值图像中间行、中间列黑白跳变次数,可将字符分为10类。进行字符识别时,首先判定待识别字符属于哪一类,然后与所在类的每一个字符的标准模板进行匹配,统计待识别字符落在每一个标准模板矩阵的不同区域的像素数,并根据不同区域的不同权值计算相似度值,相似度值最大的即为识别结果。该方法采用两级分类法对车牌字符图像进行分类,再采用基于权系数标识符矩阵的模板匹配法对车牌字符进行识别。实验结果表明,该方法提高了识别结果的准确度,对于存在字符断裂以及形状相似而容易混淆的字符有较好的识别效果。

In order to improve the accuracy of license plate recognition, a license plate character recognition method matching with a template based on weight coefficient identifier matrix is proposed. The specific method is to develop a standardized template for each license plate character before character recognition. Each standard template character is divided into three parts： pixel area, pixel edge region, non pixel and non-pixel edge region, which is taken as a criterion to generate a template matrix. According to the black-to-white jump times of number and character binary image in the middle row and middle line, the quantity of pixles which drops down on different areas of each standardized template matrix from characters under recognition is counted. The similarity value depends on the weights on the area is calculated. The maximum value is the recognition result. The method adopts secondary classification and is based on weight coefficient identifier matrix. The experiment results indicate that the algorithm is able to improve the accuracy of recognition result and has better recognition effect for the characters in which the fractures and shape similarity characters exist.