The Research Of Rapid Determination Of Phycocyanin In Spirulina Platensis Based On Spectroscopic Techniques

Phycocyanin,as a rare water-soluble blue pigment-protein complex,has a wide application prospect in food and beverage coloring,health food and other industries.The market valuation of phycocyanin is constantly improving.Spirulina platensis can synthesize phycocyanin and has been developed as a health food,with potential as a raw material for the phycocyanin industry.In this study,we studied the production capacity and optimal environmental parameters of phycocyanin in spirulina using artificial sewage conditions,and the rapid detection of phycocyanin content based on hyperspectral technology and Raman spectroscopy.The main conclusions are as follows:(1)Spirulina cultivation and phycocyanin production using wastewater containing ammonia-nitrogen and phosphorus content were studied,during which the changes of biomass and phycocyanin content of Spirulina were observed.The full-scale experiment was designed with nitrogen and phosphorus sources as the factors,and the combination of nitrogen and phosphorus source conditions most suitable for spirulina cultivation was obtained by analysis of variance,that is,NH₄Cl 0.1 g/L and K₂HPO₄0.42 g/L.Furthermore,the long-term observation proved that Spirulina can absorb the nitrogen elements in the form of ammonia nitrogen in the culture medium effectively and support growth,and the utilization rate is significantly higher than that of the nitrate nitrogen used in the normal SP medium.(2)Rapid detection of phycocyanin concentration was studied using visible-NIR and NIR band of hyperspectral imaging,and the regression models were established to realize the rapid quantitative detection.The models of visible-NIR band are better than NIR models,the optimal model is the CARS-LSSVM model of visible-NIR band.To establish regression models with better prediction ability,CARS method was introduced to extract feature variables.CARS variables were used as the input of machine learning models.The CARS-LSSVM and CARS-BPNN models of two spectral data were established respectively.Both models were very successful,and the models of visible-NIR spectral data is better than the models of NIR.The best model was the CARS-LSSVM model of visible-NIR spectra,of which R² of modeling set,validation set and prediction set were 0.999,0.989 and 0.986,respectively,and the RPD value is 8.305.In summary,the results of rapid detection of algal phycocyanin by hyperspectral imaging technology were good,and it is recommended to use visible-NIR data combined with LSSVM modeling method based on CARS characteristic band extraction.(3)Rapid detection of phycocyanin concentration was studied by confocal Raman micro-spectroscopy and a 405nm portable Raman spectrometer.The regression models of phycocyanin content were established to analyse the feasibility of quantitative detection by these two Raman technologies.Using CARS method,the CARS-LSSVM model established after extracting feature bands from data of Raman micro-spectroscopy is less successful.For the first time,we used a 405 nm portable Raman spectrometer to test the algal samples and obtain good spectral results.The best model was characteristic peaks'LSSVM model,of which R² of modeling set,validation set and prediction set were 0.908,0.908,0.907,respectively,and the RPD value is 3.357.In summary,Raman signals of phycocyanin and other pigments can be detected by confocal Raman micro-spectroscopy,but it is not suitable for quantitative modeling.The rapid detection of phycocyanin concentration using a 405nm portable Raman spectrometer is feasible,and the recommended modeling method for regression analysis is characteristic peaks'LSSVM.But further study of regression modeling and reasonable optimization of sample detection and data acquisition process is needed.