基于新模板的算法在车牌字符分割中的应用

New model of neighborhood used in segmentation of vehicle plate

针对以往车牌字符分割算法在复杂情况下分割准确率低的问题,提出了一种基于新邻域模板的连通域算法的车牌字符分割算法。采用基于新模板的连通域算法粗分割字符;通过自适应阈值投影法进行字符细分割;同时使用区域合并算法保证字符的完整性。实验结果表明,该算法将连通域法与投影法有效结合,能够同时解决汉字不连通、字符粘连和边框粘连等复杂情况,该算法对车牌字符进行分割正确率达99.5%。

Considering the low accuracy of traditional license plate character segmentation algorithm with complex situations, an optimized algorithm is proposed. It uses connected area algorithm based on templates to implement segmentation roughly. Then, adaptive threshold algorithm is introduced for further segmentation; at the same time, region merging algorithm ensures the integrity of characters. Experimental results show that unconnected problem of Chinese and character/border adhesion can be solved effectively by the combination of connected area algorithm and projection algorithm designed, the accuracy rate reaches to 99.5%.