视频图像车牌颜色辨别方法

License plate classification based on video image

在实际应用环境的视频图像中,车牌颜色极易受到天气、光照、粉尘、车牌污垢等的影响.传统的车牌颜色辨别方法主要是通过判断车牌颜色分量的阈值来实现的.在视频图像中,由于车牌颜色变化大,很难找到恰当的阈值,因而传统的车牌颜色辨别方法已不具有普遍的适用性.针对这个问题,该文提出了一种全新的基于AdaBoost算法的车牌颜色辨别方法,通过提取车牌字符和背景在RGB和HSV空间的颜色特征,训练出Ada-Boost分类器,从而对车牌颜色进行判断.实践证明,该方法不仅可以很好地判断出视频图像车牌的颜色,对于一些非车牌区域的排除也可以达到很好的效果.

In video image of practical application environment, the color of a license plate may be easily affected by weather, light, dust and dirt. Traditional methods of license plate classification are often based on the threshold of every component of a color space model. But, in video image, plate color may have a large change with different license plates, it＇s difficult to find an appropriate threshold to classify different plate colors. A new classific method based on AdaBoost algorithm was proposed in terms of the above problem. After getting the color features of plate letter and background in RGB and HSV color space model, an AdaBoost classifier was trained. Then this classifier was used to distinguish different plate colors. Practice has proven that this method can not only classify the color of a license plate, but also eliminate some none-license-plate areas.