Abstract: On the base of studying the existing license plate location algorithms, a new heuristic license plate location algorithm is proposed which is based on the statistical feature. The algorithm uses image pyramid to grade the image and then uses jumping feature to quantify the plate region’s intensive character feature. The method uses dynamic programming algorithm to calculate the statistical matrix. The range of the jumping feature which is got by prior experiments is used to select the candidate rectangles. According to the color feature, the license plate size feature, the characteristics’ number feature, etc, the final positioning result is got. A large number of experiments show that the method is accurate, efficient and can be better adapted to the environment.

Key words: heuristic license plate location algorithm, image pyramid, statistical feature, license plate location

摘要： 在研究现有车牌定位算法的基础上，提出了一种基于统计特征的启发式车牌定位算法。该算法利用图像金字塔结构将图像分级处理，将车牌区域字符密集特征量化为跳变特征，利用动态规划算法计算统计矩阵，根据事先实验得到的车牌跳变特征范围筛选统计矩阵得到候选矩形框。根据颜色特征，车牌尺寸特征，字符个数特征等筛选候选区域得到最终定位结果。大量实验表明，该方法能精确，高效地定位车牌并且对环境的适应能力比较好。

关键词: 启发式车牌定位算法, 图像金字塔, 统计特征, 车牌定位