近红外光谱结合小波神经网络鉴别苹果种类

Identification of Apple Types by Infrared Spectroscopy with Wavelet Neural Networks

将近红外光谱与小波神经网络技术结合,实现对不同种类苹果鉴别。将80个样本随机分为建模样本集和预测样本集。其中建模样本集包含60个样本,预测样本集包含20个样本。应用小波变换与主成分分析对样品数据进行预处理与特征提取。建立一个10-45-2的三层小波神经网络,实现对未知样品预测。实验结果表明,该方法对苹果的种类鉴别率达到100%,说明这种方法有很好的鉴别作用,对苹果种类的准确、无损检测具有积极的实用性。

The different types of apple were identified by the methods of wavelet neural network and infrared spectroscopy. The 80 samples were randomly divided into the training samples set and the prediction samples set. The training samples set contained 60 samples and the prediction samples set contained 20 samples. The data of the samples were pre-processed and the characteristic extracted by wavelet analysis and PCA. A 10-45-2 three layer of wavelet neural network was built to predict the unknown samples. The experimental results showed that identification rate of apple types can reach to 100%, which demonstrated that this method has good identification and positively practical application of fast and nondestructive identification of the apple types.