Abstract: License plate recognition is the key of the intelligent transport monitoring system，and the rate of license plate style identification is one of the important technical specifications in the license plate recognition system.This paper proposes a new method based on clustering and neural network to identify the license plate style.Firstly，the plate image is incline corrected，and then the effective region is extracted，finally the plate style is identified by the K-means and BP neural network.The experimental results of more than 500 images acquired under different conditions show that the method is more accurate and robust to noise and illumination variations，and the rate of identification is above 99%.

摘要： 汽车牌照自动识别系统是实现智能化道路车辆监控的基础，而车牌类型的识别率是车牌识别系统中重要的技术指标之一。提出了一种基于聚类和神经网络车牌类型识别算法。首先进行车牌的倾斜校正，其次提取车牌的有效区域，最后应用K-means边缘颜色聚类和两级BP神经网络进行车牌类型的识别。对各种条件下采集的500幅车牌图像进行实验，识别率在99%以上。实验结果表明，该算法对光照变化和噪声具有很好的鲁棒性。