LPR系统车牌定位提取方法的研究

Vehicle license plate location and segmentation in LPR system

车牌定位在车牌自动识别中起着非常重要的作用,定位准确度直接影响车牌识别的正确率。文中使用了数学形态学和几何拓扑学相结合的方法对车牌区域进行定位提取。该方法首先采用Top-Hat变换以及开、闭运算对抓拍的车辆图像进行预处理;然后采用连通体态分析法(CCA)对图像进行粗定位;最后对计算得到的车牌候选区的欧拉数进行判别,最终提取真正的车牌区域。实验证明该方法能够很好的对牌照区域顶角进行快速搜索定位,将牌照从复杂背景图像中分割出来。

Vehicle license plate locating is one of the key technoldgies for automatic License Plate Recognition （ LPR ）, and has a great impact on character segmentation and character recognition. In this paper, morphology and topology are used in plate location. Image preprocessing is first adopted,which is implemented by Top-Hat transform,opening and closing operations. Connected components analysis is then performed to detect candidate regions. At last,the real license plate area is obtained according to the Euler number of candidate regions. Experiments show that it can realize the fast search, and location of the corner angles of the license plate and segment the license plate from the complex image.