一种基于机器学习的车牌识别系统的设计

以车牌识别的实用性为目的,设计一种鲁棒的车牌识别系统。首先提出了Sobel-Color算法,以Sobel边缘和颜色两种特征进行车牌定位,并结合MSER算法,设计了一种可靠的车牌定位方法来获取候选车牌区域,然后采用SVM算法对候选车牌区域进行车牌判断;最后根据车牌特征设计了一种车牌字符分割算法,能正确分割车牌的各个字符,并有效地去除车牌边缘部分的虚假字符,又根据分割出的车牌字符特征对LeNet-5深度网络模型进行改进,然后采用改进的LeNet-5网络对车牌字符进行识别。对设计的车牌识别系统进行了正常条件测试、恶劣条件测试以及效率测试等实验,实验结果表明设计的车牌定位和车牌判断方法具有较高的可靠性,车牌字符识别具有较高的准确率,因而设计的车牌识别系统具有较好的鲁棒性和实用性。

Design of License Plate Recognition System Based on Machine Learning

Aiming at the practicability of license plate recognition,a robust license plate recognition system is designed. Firstly,Sobel-Color algorithm is proposed to locate license plate based on Sobel edge and color features,and combined with MSER algorithm,a reliable license plate location method is designed to obtain candidate license plate regions,and then the SVM algorithm is used to judge them. Finally,a license plate character segmentation algorithm is designed according to the license plate characteristics,which can segment the characters of the license plate correctly,and effectively remove the false characters of the edge of the license plate. According to the characteristics of the license plate characters,the LeNet-5 depth network model is improved,which is used to recognize the license plate characters. The normal condition test,harsh condition test and efficiency test of the license plate recognition system are carried out. The experiment shows that the method of license plate location and license plate judgment has high reliability,and the license plate character recognition has high accuracy. Therefore,the designed license plate recognition system has better robustness and practicability.