基于字符包围盒特征参数的车牌定位分割研究

现有的车牌识别流程中,车牌定位和字符分割关联性较小,有效融合这两个过程可以提高检测的效率。为此,提出了一种基于字符包围盒特征参数的车牌字符定位分割算法。该方法首先使用高斯-拉普拉斯算子对车牌图像进行滤波和边缘检测,在此基础上,使用连通区域分析和轴对齐包围盒检测和表征边缘图像中的连通域,根据车牌字符的分布规律来筛选目标包围盒,最后基于目标包围盒中心点对车牌字符进行倾斜矫正和定位分割。实验结果表明,该方法可以有效地定位车牌和分割字符,准确率可达到93.5%。

REAEARCH OF LICENSE PLATE LOCALIZATION AND SEGMENTATION BASED ONCHARACTER BOUNDINGBOX PARAMETER

In the existing algorithms of license plate recognition, there is less relationship between plate localization and character segmentation. Effective integration of these two processes can improve the detection efficiency during the process of license plate recognition. To this end, a license plate localization and segmentation method is proposed based on character bounding box parameter. Firstly, Laplacian of Gaussian is used to filter image and detect edge. Then CCA and AAB bounding box are used to detect and label the connected components. Target bounding boxes are selected due to the distribution law of the real character. Finally, the center points of the target bounding boxes are utilized for the final localization, tilt correction and segmentation. Simulation results show that the proposed method can effectively locate the plate and split the character, with accuracy up to 93.5%.