摘要
本文提出了一种基于形态学和分形几何的车牌字符分割算法,首先采用差分盒维数算法估算车牌图像的分形维数并进行二值化处理,然后运用数学形态学方法突出车牌字符特征;最后结合先验知识对分割出的字符进行筛选,从而得到正确的车牌分割字符。该算法有效地提取了车牌字符信息,具有良好的分割效果。
This paper puts forward a vehicle license plate character segmentation method based on mathematical morphology and fractal geometry. Firstly, it adapts the differential box-counting dimension to estimate the fractal dimension of license plate region, and make the regional characteristics of license plate outstanding combined with morphology. Secondly, it removes up-down-direction boundary of license plate image by using horizontal projection, and divides the image into single characters by using vertical projection. At last, it screens divided characters combined with the prior knowledge such as intrinsic characteristics of license plate, then get the correct divided characters. This algorithm extracts information of license plate characters effectively and has a good segmentation effect.
引文
[1]董慧颖,曹仁帅.汽车牌照自动识别系统中字符分割算法研究[J].沈阳工业学院学报,2003,22(4):1-4.
[2]潘中杰,谭洪舟.模板匹配法和垂直投影法相结合的一种新的车牌字符分割方法[J].自动化与信息工程,2007,2:11-13.
[3]刘明军,谢宏霖,孙雪松,等.车牌字符分割算法的比较研究[J].济南大学学报(自然科学版),2006,20(3):245-248.
[4]MandelbrotBB.The fractal geometry of nature[M].California:Freeman,1982:174-195.
[5]王东生,曹磊.混沌、分形及其应用[M].合肥:中国科技大学出版社,1995:282-290.