多层前馈模糊神经网络进行图像识别

A PATTERN RECOGNITION METHOD BASED ON THE MULTILAYER FEED - FORWARD FUZZY NEURAL NETWORK

神经网络和模糊技术在模式识别领域中已有了广泛应用,两者有着各自的优势。针对神经网络模式识别中所遇到的问题,为了进一步提高分类器在样本分布不清晰情况下的识别能力,本文提出了两种将模糊机制引入神经网络的方法——输入模糊化方法和隐层模糊化方法,并在此基础上分别构造了模糊神经网络。实验结果表明,模糊神经网络较好地结合了神经网络和模糊技术的优点,取得了比传统神经网络更好的识别结果。

The artificial neural network and the fuzzy algorithm have been widely used for pattern recognition . In order to utilize both of their advantages, we propose two kinds of fuzzy neural network by introducing the fuzzification into the input layer and the hidden layer of the neural network. Parts recognition experiments show especially promising results in classification of non - sparse or overlapping patterns.