基于边缘分析和颜色统计的车牌精确定位新方法*

针对车牌的边界定位不准确和伪车牌较多的问题,提出了一种边缘分析和颜色统计相结合的车牌精确定位的新框架。该框架主要分为预处理、粗定位、精确定位和伪车牌排除四个模块。对图像作边缘检测和二值化等预处理,用投影法粗定位出候选区域,利用候选区域及其周围的边缘和颜色的信息实现车牌的精确定位。对于定位结果有多个候选区域的情况,对候选区域进行排序,再将排序结果按顺序传入字符分割模块,从而有效排除非车牌区域的影响。实验结果表明,该方法精确度和准确率高、实时性强,适用于不同的应用场合。

New approach for precise license plate locating based on edge analysis and color statistics

For the problem of the boundaries of the license localization being not correct and existing many non-license plate regions, the paper proposed a new framework of precise license plate localization combined edge analysis and color statistics. The framework included four parts, preprocessing, coarse positioning, precise positioning and non-license plate rejection. Firstly, this method carried out preprocessing such as edge detection and binarization, then adapted the method of projection to located some candidate regions coarsely, after that, analysed edge and color information in and around candidate region to rea-lize the precise locating. For the situation of existing more than one candidate regions, the method sorted the candidate regions, then passed the sorted regions to the character segmentation module, it could effectively eliminate the influence of non-license plate.Experiment results show that the method is highly accurate and precise, high real-time, adapting to different situations.