摘要
车牌定位是车牌识别系统的重要组成部分,目前常用的车牌定位方法主要受环境尤其是光照影响较大。针对这一情况,提出基于最大稳定极值区域特征的车牌定位算法,利用最大稳定极值区域特征特有的仿射不变性和对光照的适应性,提取图像中最大稳定极值区域;尤其是车牌字符区域。在排除部分噪声区域后,根据车牌字符区域稳定的几何特征和排列规则,将满足条件的相邻字符区域组成最近邻对,进一步剔除噪声区域。然后将所提取的最近邻对进行合并即可以得到所有可能的车牌区域。实验结果表明,相比较目前常用车牌字符切分算法,在切分的准确性和稳定性上都有较大提高。
License plate localization is a important step in the License Plate Recognition system. Nowadays,the main localization method depends on the environment especially the light have agreat effect on the detection accuracy. In response to this situation,a license plate detection algorithm was proposed based on maximum stable extremal regions. The main idea is using MSER＇s affine invariant features and MSER＇s invariance to lighting change to extract license plate image＇s MSERS,especially the license plate character regions. After exckuding some noise regions,then according to the geometrical characteristics and spatial arrangement of license plate characters,the adjacent characters can construct a nearest neighbor pairs if the two characters reions meet the corresponding conditions,which also can exclude interference regions. Then all the extracted nearest neighbor pairs was merged,in this way can be detect all the candidate license plate regions. Experimental result shows that the proposed algorithm improve the the accuracy and stability compared to the current license plate detection algorithm.
出处
《科学技术与工程》
北大核心
2015年第31期212-217,共6页
Science Technology and Engineering
