基于多尺度模板匹配和部件模型的车牌字符分割方法

为提高车牌字符分割的准确率,提出了一种基于多尺度模板匹配和部件模型的车牌字符分割方法. 对单层车牌,根据车牌结构特征建立多尺度模板作用于车牌候选区域,通过投影得分估计出最佳模板对应的尺度和位置信息;对双层车牌,使用部件模型对双层车牌进行建模,双层车牌的上层字符区域和下层字符区域分别对应部件模型的一个部件,通过多尺度的模板匹配得到上下层部件的候选集合,利用部件模型中部件之间的几何约束得到最终的车牌字符分割结果. 实验结果表明,所提出的方法可以有效进行单/双层车牌的字符分割. 

License Plate Character Segmentation Based on Multiple Scale Templates Matching and Part-Based Model

A license plate character segmentation method based on multiple scale templates matching and part-based model was proposed to improve the precision of license plate character segmentation. Characters of single layer plates were extracted by sliding the multiple scale templates, which were designed according to the structures of the plates, in the plate candidate area. The double layer plates were modeled by part-based model, the top layer and the bottom layer are the two parts in the model. Candidates of the two parts were obtained by the multiple scale templates matching, and the geometrical constraint between the two parts was used to get the final segmentation result of the characters. The experiment results indicate that the proposed approach can segment characters for both single layer plates and double layer plates effectively.