基于Pseudo-Zernike不变矩的PNN车牌汉字识别  

PNN Recognition of License Plate Chinese Characters Based on Pseudo-Zernike Invariant Moments

作　　者：高全华[1] 王晋国[1] 孙锋利[2] 

机构地区：[1]长安大学理学院,西安710064 [2]西北工业大学电子信息学院,西安710077

出　　处：《计算机工程》2009年第4期196-198,共3页Computer Engineering

摘　　要：基于不变矩理论,提出一种应用概率神经网络作为识别器的车牌汉字识别技术。利用Pseudo-Zernike矩特征的旋转不变性和良好的抗噪性能,将其作为车牌汉字识别的特征矢量,结合Pseudo-Zernike矩的快速算法和概率神经网络识别器快速学习和识别的性能,可适应实时环境下所获取的车牌汉字灰度图像的识别,具有较高的准确率,实验结果表明了该方法的有效性。This paper presents a novel approach based on Pseudo-Zernike Invariant Moments（PZIM） and Probabilistic Neural Network（PNN） to recognize license plate Chinese characters. The approach makes better use of the rotation invariant and good anti-noise performance of Pseudo-Zernike moments and quick learning rate of PNN, and thus provides a real-time recognition of gray character images by utilizing Pseudo-Zernike moments as feature vectors and Probabilistic Neural Network as classifier. Numeral experiment confirms that it is an effective way to classify license plate Chinese characters.

关 键 词：车牌识别 Pseudo—Zernike不变矩 概率神经网络 

分 类 号：TP389[自动化与计算机技术—计算机系统结构]