基于DSP的图像处理在车牌识别中的应用

The Application of The Image Process Based on DSP in the LPR

车牌识别模块是车牌识别(LPR)系统的核心。论文根据国内汽车牌照的特点,对车牌识别模块中的预处理、字符分割及字符识别技术提出了改进的算法,并基于DSP实现了对车牌纯字符区域的准确提取、分割。改进点有采用对边缘锐化后的二值图像进行局部投影去除车牌背景、对各字符的外部轮廓进行统计特征提取以及充分利用数字"1"自身的特点设计识别方案。通过Code Composer Studio(CCS)对358副车牌图像进行了仿真测试,识别率为99.16%。

The subsystem of License plate recognition is the core of the license plate recognition （LPR）. According to the characteristics of the domestic automobile license, this paper proposes an improved algorithm of the subsystem including pretreatment of license plate image, license plate character segment and character recognition. Based on DSP, the precise extraction and segmentation of the character area from the license plate are achieved. In this paper, the background of the license plate is discarded by local projection of the binary image after the edge detection, the external outline of each character is extracted by the statistic result, and the template matching is on basis of the feature vector; the specialty of ＂1＂ is fully used to discriminate ＂1＂ from other numbers. The performance of the system has been investigated on real located images of about 358 license plates extracted under various illumination conditions in the CCS. Recognition of about 99.16% shows that the system is quite efficient.