摘要
基于伪zernike矩的不变性,提出了基于伪zernike矩特征提取的车牌字符识别方法。在车牌区域定位的基础上,通过对图像的二值化过程和字符图像分割等一系列处理,在进行归一化处理后分别提取伪zernike的高阶矩。将提取的伪zernike矩作为字符的特征描述输入到BP神经网络进行训练,最后进行车牌字符分类识别。通过实验证明了该方法的的可行性。测试结果表明,这种方法实用有效,识别效果优于HU矩和zernike矩。而且可以计算出错误率和可识别的最佳矩,减小了计算量且增强了字符识别的实时性。
Based on the invariance of the Pseudo-zernike Moments,this paper presents a novel method of lisence plate character recognition based on Pseudo-zernike Moments(PZM).On the basis of plate area having been positived,affter a serisr of processing including binarization,segmengtion,normalization we extracted highlevel Peudo-Prnike Mments these we take as the description of character feature to make the BP neural network training.Finally the license plate characters can be classified and recognised.The feasibility of the method can be proved by the experiments.Expriments results show that the presented approach achieves better recognition accuracy than using HU moments and zernike moment.And it also can calculate the error-rate and the optimal number of moments of the recognition,decreasing the amount of computation and enhancing the real-time while recognising the character.
出处
《电子测试》
2012年第8期19-23,共5页
Electronic Test
关键词
车牌字符识别
伪ZERNIKE矩
BP神经网络
特征提取
lisence plate character recognition
Pseudo-zernike moments
the BP neural network
feathure extraction