摘要
目前,车牌识别系统大都采用单摄像头获取车牌图像,考虑到这种方法对于污染严重、光照不佳的车牌识别率较低的缺点,提出了一种基于像素级图像融合的车牌识别算法;首先采集不同摄像头的多角度车牌图像,定位分割后进行图像配准,然后利用小波图像融合算法进行图像融合,得到新的车牌图像;此方法综合利用了多源图像的互补信息,提高了识别的置信度;通过多组不同条件车牌的对比实验,与传统车牌识别方法以及决策级车牌融合方法相比,此方法能有效提高车牌的识别速度及识别率。
Currently, most of the license plate recognition systems use single camera to obtain license plate images which have the disad- vantage of low recognition rate in serious pollution and poor lighting conditions of license plates. This paper proposes a license plate recogni- tion method based on the pixel--level image fusion technology： first, collect the multi--angle license plate images from different cameras, then images registration after localization and segmentation, image fusion using wavelet algorithm to get a new license plate image. This method of comprehensive utilization of the complementary information of multi source images improves the recognition confidence. By com- parative experiments for many license plates of different conditions, this method can effectively improve the license plate recognition speed and recognition rate compared with the traditional recognition method as well as the decision level image fusion technology.
出处
《计算机测量与控制》
北大核心
2013年第3期791-793,共3页
Computer Measurement &Control
基金
国家自然科学基金(61071175)
关键词
像素级
图像融合
车牌识别
pixel--level
image fusion
license plate recognition
