摘要： 提出一种使用模糊神经网络对纸币新旧程度进行实时分类的方法。为了达到实时性要求，该方法使用图像区域的一阶矩作为纸币新旧分类的特征，使用模糊神经网络作为分类器。在分类器的隶属函数生成层对特征向量向目标空间进行映射，在网络的推理层对纸币的新旧进行分析，在去模糊化层给出纸币新旧的定量分析结果。实验结果表明，该方法对纸币新旧的分类是准确和稳定的。

关键词: 特征提取, 图像处理, 模糊神经网络, 纸币清分

Abstract: A real-time banknote sorting method is proposed which uses fuzzy neural networks. In order to satisfy the requirement of real-time, the first-order moments of the pixels values in the local blocks of the image are used as the feature vector of the banknote. The fuzzy neural networks is used to classify the new-old degree of the banknote. The membership function in the network maps the feature space to the target space. The inference layer of the networks analyzes the new-old degree of the banknote. The decision of quality is given by the defuzzy layer. Experimental results show that the method is able to classify the banknote accurately and steadily.

Key words: features extraction, image processing, fuzzy neural networks, banknote sorting

中图分类号: 

• TP391