车辆牌照图形的分割是车牌识别(VLPR)中的关键技术。对于通常采用的基于二值化结果的分割算法,车辆牌照图像的质量往往不佳,二值化后将会产生大量噪声,从而影响车牌分割,使系统整体识别率不高。文中提出了适合于户外流动车辆车牌分割的彩色车牌峰谷分割算法(CLCSA),不需对车牌进行二值化,通过对HIS空间的H分量和RGB空间的B分量适当调整,得到一个峰谷值差异较大的投影图像,从而为车牌正确切分提供了理想的位置。实验结果表明,蓝车牌和黄车牌的正确切分率达到97.5%,明显优于传统的对二值化图像的处理,特别是对于斜车牌和模糊车牌切分效果更加明显,更适用于实际应用。
Character segmentation for vehicle license plates is one of the critical techniques for the intelligent transportation system(ITS). Vehicle license plate segmentation algorithms are usually based on binarization. The quality of vehicle license plate image is always very poor. Through binarizing these images much noise is produced which will reduce the whole recognition rate. So a new segmentation algorithm for flowing plates outdoors---CLCSA (coloured license - plate crest segmentation arithmetic without binarization) is presented. CLCSA algorithm needn't binarizaion. Through adjusting the H of the HIS space and B of the RGB space, a projection with obviously distinct can be Rotten, which implies the correct segmentation position. Experiments show that CLCSA is better than traditional techniques, with 97.5% correct segmentation rate of blue and yellow plates. CLCSA is better in practlcal,especially for inclined or blur plates.