一种改进的夜间车牌定位和字符分割算法

针对传统的车牌定位算法对夜间车牌图像定位效果差的缺点，提出了基于支持向量机的定位方法；针对夜间车牌图像的灰度分布不均匀的缺点，采用局部二值拟合模型方法进行字符分割。实验结果证明该方法定位精确，对于包含车头灯或车尾灯的夜间图像，定位方法均有效，并且提供了准确的光滑闭合边界，精度可以达到亚像素级，系统的识别准确率高。

An Improved Agorithm of Night Plate Location and Character Segmentation

The traditional number plate location method is not good in the night. Aiming at this shortcoming, a method for number plate location based on SVM is proposed. Then aiming at the gray uneven distribution in the night vehicle images, number binary fitting model that presented by Li, etc, is used to segment characters. Experimental results show that the method locates accurately, and is effective for the nighttime images involve the headlights or the taillights, and provides the smooth and closing boundary accurately, gives the precision which can reach subpixel level. The recognition quality of system is as good as what we expected.