基于SVM的车牌自动识别系统

A Recognition System for License Plate Based on SVM

为了解决识别率和识别速度同时提高的难点,将支持向量机(SVM)算法应用于车牌字符识别。通过与无字符特征提取的BP神经网络相比较,在小样本的情况下,该方法的识别率远优于神经网络,可很好地解决神经网络方法中收敛速度慢、易陷入局部极小的问题。实验结果表明:在训练样本较少且无字符特征提取的情况下,此系统具有较高的识别率和识别速度,并具有很好的分类推广能力。

To solve the problem of high recognition rate and recognition of license plate. Comparing with BP network, this speed, support vector machine （SVM） will be used in the character method works better in improving recognition rate and speed and avoiding the problem of the local optimal solution of BP network and it needs no empirical knowledge. Experiment results show that the system has high recognition rate and speed which can be extended greatly.