摘要
针对车牌字符识别中大部分单一特征提取方法在字符识别上的局限性,提出了一种车牌字符多特征提取方法。在经过预处理后的车牌细化字符基础上提取字符4个侧面的笔画特征、拐点特征、轮廓累积特征及字符内部像素特征,构建出一个维度较低的特征向量集,然后分别采用支持向量机、K近邻算法、BP神经网络、径向基神经网络对陆丰高速公路实地拍摄的车牌图片进行测试并分别与模板匹配方法、网格法、基于小波矩方法比较,实验结果表明提出的车牌字符多特征提取方法识别率高,鲁棒性好。
To solve the limitation of most of the single-feature extraction methods in vehicle license plate recognition,a method based on multi-feature extraction is presented.After pre-processing,different kinds of features are extracted,including the strokes features,inflection point features and contour features of four outsides of characters as well as the internal pixel features based on the thinned characters.These features are then converted into a lower-dimension feature vector set,on which Support Vector Machines(SVM),K Nearest Neighbor(KNN),Back Propagation Neural Network(BP-NN),Radial Basis Function Neural Network(RBF-NN) can be built.These classifiers are tested on the vehicle images taken from the Lufeng Freeway.This paper compares the proposed method with the pattern matching method,grid method and the wavelet moment method on performance.The experimental results show that the proposed multi-feature extraction method has high recognition rate and robustness.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第23期228-231,共4页
Computer Engineering and Applications
关键词
车牌字符识别
多特征提取
支持向量机
神经网络
K近邻
vehicle license plate recognition
multiple features extraction
support vector machine
neural network
K Nearest Neighbo(rKNN)