Identification Technology for Three-Dimensional Fluorescence Spectrum of Mineral Oil Based on Lifting Wavelet - Multi-Resolution Orthogonal Multi-Wavelet Network

The fluorescence characteristic spectrums of samples obtained from the parameterization for three-dimensional fluorescence spectrum of mineral oil are diverse, and therefore, the species identification of mineral oil is difficult to be realized through simple formulas when the species identification of mineral oil is conducted by means of three-dimensional fluorescence spectrum technology. In this paper, lifting wavelet is adopted to conduct noise reduction on fluorescence spectrum signal of mineral oil extracted by spectrometer firstly, and then a hierarchical multi-resolution multi-wavelet neural network with the feature of localized learning is adopted to realize the classified recognition for fluorescence spectrum of mineral oil. The experiment shows that, the network not only maintains all the advantages of wavelet neural network, but also possesses better approximation properties than uni-wavelet neural network, which realizes the identification of mineral oil by lower frequency of training and of which the identification accuracy is up to 95%.