Fluorescence spectrum recognition of pesticides based on wavelet neural network

The fluorescence spectrum of pesticides whose structures are very similar overlap in a certain wavelength range. To the classification and recognition of overlapping fluorescence spectrum, BP network has the shortcomings of slow training speed and high error rate. An improved wavelet neural network (WNN) is presented in this paper. The network topology is given, wavelet basis is selected and its network algorithm is designed to carry out the design of experimental system. By using the WNN and BP network separately, the simulation research of fluorescence spectrum classification of carbofuran and carbaryl has been done. The results show that WNN has a higher resolution. To minor structural differences of spectrum, it has a better recognition capability and higher measuring accuracy.