The Composition Analysis Of The Complexed Mixed Solution Based On The Mutil-dimentional And Mutil-mode Hyperspectral System

Complex mixed solutions have been widely used in the fields of food,medicines, environment and agriculture, etc. Its application is concernsboth the quality and production efficiency of the industrial andagricultural products, and also concerns people’s livelihood, such as foodsafety. Modern industrial and agricultural production requires rapid,non-destructive, online detection method for monitoring and detecting ofindustrial and agricultural products in order to ensure stability of thequality and safety of food, medicine and other products.Most of the liquid food and agricultural products are cloudy or opaque inthe visible and near-infrared regions. Light scattering caused by particlesleads to poor repeatability, low signal-to-noise ratio (SNR) and poorunderstanding of the physical mechanism of the VIS-NIR spectroscopy.In this paper, we proposed a method called mutil-dimentional and mutilmode hyperspectral system combining the benefits of absorption andscattering properties for analyzing the chemical compositions of thecomplex mixed solution.First, to verify the feasibility of the analysis of complex mixed solutionwith the mutil-dimentional and mutil mode hyperspectral technique, wetook Monte Carlo method to simulate the distribution of the diffuse lightof the Intralipid-Ink model at the range of650-1100nm. The distributionof the diffuse light was obviously different at different wavelengths. Itwas proved that the mutil-dimentional and mutil mode hyperspectraltechnique can use absorbtion and scattering characteristic changes withthe wavelengths of the analyte, and greatly improved the signal to noiseratio, it has the potential to significantly enhance the ability of thecomponent analysis of complex mixed solutions.Then, a mutil-dimentional and mutil-mode hyperspectral acquisitiondevice was designed to collect the diffuse reflectance hyperspectralimages of the analytes (milk and fruit juice from different manufacturers). The hyperspectral images showed that the differences between differentsamples was obvious than that of traditional spectrum. This researchreveals that the hyperspectral technology is feasible in complex mixedsolution component analysis.At last, we proposed a method for component analysis of complex mixedsolutions based on multi-dimensional diffuse reflectance/transmissionspectroscopy. A set of instrument was designed of light source,translation stage and the spectrometer, and was used to aquire the diffusereflectance/transmission spectrum signal at multi-points in ratial direction.The PLS and N-PLS were taken to analyze the information carried by3-dimensional spectrum signals for predicting the components of theanalyte. The results showed that predicting accuracy of strong scatteringand absorption material were increasing with the growing number of thesample point on the surface of the diffuse reflectance/transmissionmethod, the optimized number of sample points was figured out withdifferent phaotom, and the predicting accuracy of the multi-way partialleast squares (N-PLS) was higher than that of the partial least squaresmethod (PLS).It is proved that, by the way of improving the the spectrums collectedfrom different points can raise the signal to noise radio of the spectrums,comparing with the traditional spectrum technology using a single opticalproperty such as absorbance or reflectance, this method had consideredboth the impact of absorption and scattering characteristics of the analyte.So the use of a variety of optical properties of the analytes can make animprovement of the accuracy of the modeling and forecasting model, andalso provide a basis for component analysis of the complex mixedsolution base on multi-dimensional diffuse reflectance spectroscopy.