基于场景分类及灰度跳变的车牌定位方法

车牌定位是自动车牌识别系统实现的关键。提出一种基于场景分类及灰度跳变的车牌定位算法。该算法对彩色图像进行场景分析,将图像分类为白天场景类或夜晚场景类。这两类场景的字符与背景的灰度跳变值不同,一般白天场景类的灰度跳变值较大,夜晚场景类的灰度跳变值较小。利用不同的灰度跳变值快速提取出几块车牌候选区域,对不同的场景用不同的方法最终选取一块区域。实验结果显示本文提出的方法对图像场景分类准确率达到98.2%,车牌定位的准确率达到98.5%。

License Plate Location Based on Scene Classification and Gray Level Jump

License plate location plays an important role in automatic license plate recognition system. This paper proposed a new license plate location algorithm, which based on scene classification and gray level jump. Firstly, the original color images are classified into two scenes, night and day, which have different gray level jump values of license character and background. Generally the day value is bigger than the night one. Secondly, several elective zones are segmented from the gray image according to the gray level values. Finally, only one region is located from the elective zones by respective method in the two scenes,. Experimental results show that the accuracy of scene classification is 98.2% and that of license plate location is 98.5%.