强噪声车牌图像中的字符识别

当字符图像受到噪声污染，甚至发生缺损、污染时，会增加车牌识别系统（LPR）准确识别的难度。在图像处理方面，提出了一种改进的全局动态二值化法，即有效减弱了光照不均的影响。又保证了较快的处理速度。此外，直接对二值图像进行噪声滤除比灰度图像的噪声滤除响应速度更快。在文字识别方面，将处理好的字符图像经过模板匹配进行初次分类，然后将图像分成4个子区域并赋予不同的权值。通过特征区权值模板进行精确判别，以提高车牌字符的整体识别率。实验数据分析表明，该识别方法有很强的抗噪声能力，且识别速度较快。

Character Recognition of Plate License with Strong Noise

The recognition of License Plate Recognition(LPR) will be more difficult when plate license character image is stained or fractured.As to image processing,an improved whole range dynamic binaryzation algorithm is issued in this lecture,which can not only reduce the influence of uneven illumination,but also keep higher processing speed.Meanwhile noise reduction is directly operated on binary image,which is faster than that on gray image.In order to keep high recognition rate,the character is sorted in first level by template matching,then divided into 4 sub-regions and make the exact recognition by characteristic regional template matching.From the data of trial,this system has strong antinoise ability and high recognition speed.