基于评分模型的车牌字符识别方法

摘要

现如今摄像头的普遍应用,使运用于监控视频中的车牌识别算法越来越受到重视. 从监控视频中切割出来的车牌字符图片往往具有低像素、模糊、字符混淆严重或者边缘凹凸不齐等缺陷,针对传统的车牌字符识别算法对于这种类型车牌字符的低识别率,提出一种基于评分模型的车牌字符识别算法. 该算法是将字符图片分成若干个方格,全部格子根据白点的数目划分为不同的等级,标准字符模板也是如此划分等级. 在识别匹配时,待识别字符图片与字符模板相对应的方格分别进行等级匹配,按照该算法中的评分模型打出相应的分数,有加分、减分或者给零分,最后将分数相加,匹配得出最高分的字符模板将是最终的识别结果. 实验结果验证了该算法的有效性.%With the wide application of camera nowadays, the license plate recognition algorithm used in monitoring video attracts more and more attentions.The images of license plate character cutting from monitoring video usually have defects such as low pixel, blurring, serious character confusion or jagged edge.Traditional license plate character recognition algorithm has low recognition rate on the plate characters of such type.In light of this, we propose a scoring model-based recognition method for license plate characters.The algorithm divides the character image into a couple of grids, and all the grids are then divided into different grades according to the number of the white points in them, and the grades of standard character templates are so divided as well.When recognising and matching, the character images to be recognised will be carried out the grade matching with the grids corresponding to character template separately, and will be marked the correlated scores in accordance with the scoring model in the algorithm, including score gain, decrease or zero, and at last the scores are added up, the character template with highest score derived from matching will be the final recognition result.Experimental results have verified the effectiveness of the algorithm.