基于伴生与互补颜色特征的车牌字符分割新方法

New method the vehicle license plate character segmentation based on concomitant and complementary color features

针对车牌区域的伴生与互补的颜色特征,提出一套字符分割的新算法.该算法利用定位到的车牌区域的原始颜色特征,直接进行字符及背景的提取,完成2值化过程,避免了图像处理过程中带来的误差,因此产生的噪声较少;最后利用垂直积分投影进行字符分割.对200幅从交通卡口获取的真实的彩色图像进行试验,算法对车牌破损,污垢、铆钉及字符粘连、过饱和等造成的影响均不敏感,其准确性,鲁棒性及实用性均较好.

A new method for character segmentation is given depending on the concomitant and complementary color features in the license plate area. The character color areas and background color areas are directly extracted from the original image of the license plate area. This was finished by binary process. By these methods, the errors arising during the image processing can be avoided, and lessen the noise made. Finally, the characters are segmented by the vertical projection method. The experiments on 200 automobile real RGB color images, which are taken under various conditions from traffic stations, show high accuracy. Robustness, practicability and lower sensitivity to the effects of plate disrepair, dirt, rivet and characters juncture can be obtained.