摘要
车牌定位是车牌自动识别系统中的关键技术之一。本文采用多层次分割的思想 ,每次分割都尽可能地减少分析范围 ,经过多次分割后最终定位出车牌区域。第一次分割采用固定分块划分出若干大小相同的特征区域 ,在各区域内通过水平差分突出垂直方向纹理特征 ,寻找水平差分累加值最大的区域即为车牌的大致位置 ;第二次分割主要对穿过字符区域的特定位置处的两条直线选取最佳阈值进行二值化处理 ,再利用正态分布的一些性质分割出车牌左右边界 ,最后一次分割过程中采用投影法 ,同样利用正态分布性质进行分割 ,找出车牌上下边界。实验表明本算法简洁 ,显著地提高了车牌定位速度。
License plate location is the key among the whole process of Car License Plate Recognition System. In this paper, the thinking on poly-hiberarchy is adopted to reduce the scope of analyses during every step of license plate segmentation, so the ultimate location result is segmented out throughout many times of segmentations. In the first segmtation, the whole image is curved up some number of blocks with the same size,and then vertial texture features are enhanced by the means of the difference in the horizontal direction,and finally in the way of seeking the maximum among difference accumulating values in the blocks,approximate region of license plate is found. In the second segmentation, the two lines with satisfaction of some conditions are binarized,which are selected from approximate region of license plate.The right and left borderies are segemented by making use of some properties of normal school.At the last segmentation,the projection method is adopted,and some properties of normal school are depended on for seeking the top and bottom borderies of license plate. The results show that this algorithm is simple and easy, and distinctly shortens the time of location.
出处
《中国体视学与图像分析》
2004年第4期239-243,共5页
Chinese Journal of Stereology and Image Analysis
关键词
多层次
阈值
二值化
正态分布
poly-hiberarchy
thresholding
binarization
normal distribution
