基于小波神经网络的农药荧光光谱识别  

Fluorescence Spectral Recognition of Pesticides Based on Wavelet Neural Network

作　　者：王玉田[1] 李长吾[1] 李艳春[1] 张雷[1] 

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004

出　　处：《计量学报》2008年第1期84-86,91,共4页Acta Metrologica Sinica

基　　金：国家自然科学基金(60272027)

摘　　要：有些农药结构相似,它们的荧光光谱在很大波长范围内相互重叠。多组分混合时,传统的荧光光谱分析法很难对其进行分类识别。介绍了一种基于小波分析构造的新型神经网络——小波神经网络,利用它并适当选取网络结构和小波基,实现了对卡死克、盖虫散和吡虫啉三种农药荧光光谱的分类识别。在小波神经网络中,采用Morlet母小波和一维搜索变步长共轭梯度优化方法。实验表明,小波神经网络对光谱间的细微结构差别具有良好的识别能力。通过比较发现,在分类识别方面,小波神经网络比BP网络具有更高的分辨率及较少的训练次数。For the pesticides with similar structures, their fluorescence spectra overlap in a wide wavelength range, For the polycomponent mixture, the conventional fluorescence spectrum analysis method hardly can identify them, A new type of neural network wavelet neural network is introduced, which is constructed based on wavelet analysis. The classification of flufenoxuron, hexaflumuron and imidacloprld are realized with adaptive network structure and wavelet basis, in the wavelet neural network, Morlet mother wavelet and line search conjugate gradient optimization method are used, The experiment results show that wavelet neural network has the better ability to identify the fine structure difference between the spectra. Compared with BP networks, wavelet neural network has higher resolution and less training times.

关 键 词：计量学 农药 荧光光谱 光谱识别 小波神经网络 

分 类 号：TB99[一般工业技术—计量学]