Abstract: 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.

Key words: license plate recognition, Pseudo-Zernike Invariant Moments(PZIM), Probabilistic Neural Network(PNN)

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。

关键词: 车牌识别, Pseudo-Zernike不变矩, 概率神经网络

CLC Number: 

• TP389