| In the world, there are a lot of species of microalgae, which are rich in high-value nutrients and chemicals like carotenoids, lipids, proteins, etc. Most of of microalgal extracts have important applications in different fields, such as medical industry, food industy, renewable energy and others. Thus, in order to help reduce costs and increase production for microalgal propagation process or microalgal industrialization, it’s necessary to study how to rapidly and precisely monitor the internal and external information of microalgal liquids. In this work, Vis/NIR spectroscopy, hyperspectral imaging technology and Raman microspectroscopy combined with chemometrics were used for rapid indentificaiton of different species of microalgae and determination of the internal information (primarily pigments and lipids) of microalgal liquids and cells in vivo. The main research contents and results are shown as follow:(1) Immersive Vis/NIR spectroscopy and Raman microspectroscopy were applied to identify 5 different species of microalgae with the established discriminative model. Totally 150 samples of 5 species of mircroalgae were collected by a self built immersive Vis/NIR spectroscopy. Then the models based on different preprocessing algorithms, different methods of effective wavelength selection including x-LW and SPA and different modeling approaches of PLS, SVM, LDA and ELM were compared by accuracy of prediction set, the result showed that SPA-LDA, SPA-ELM and x-LW-ELM models reached the best effects and their prediction accuracy all got 100%. In addition, another spectrums of 150 samples were collected by Raman microspectroscopy, and the fluorescences of which were substracted by RCF algorithms. A method of maximal peaks ratio was proposed to adjust different situations of same species of microalgae that spectral change influced by collecting over time of same cell and different lifecycle of the microalgae cells. Then 10 effective wavelengths in connection with pigments were selected to use PLS, SVM, LDA and ELM to build the model, finally the ELM model got the best prediction accuracy of 100%. The study above showed that these two spectral methods could effectively and fastly identify these 5 species of microalgae.(2) Immersive Vis/NIR spectroscopy, reflectional and transmissional hyperspectral imaging technology were applied to monitor the chloropyll a, chloropyll b and carotenoid content per unit volume with quantitative models. Different preprocessing methods and modeling algorithms of PLS, MLR and LS-SVM were used to compare Rpre and RPD of prediction sets. When using immersive Vis/NIR spectroscopy, the best coefficients of determination of prediction (Rpre/RPD) values of the models were 0.9684/2.9461,0.9146/1.9477 and 0.9638/3.0083 for spectral analysis of the chloropyll a, chloropyll b and carotenoid content per unit volume; When using transmissional hyperspectral imaging technology, the best coefficients of determination of prediction (Rpre/RPD) values of the models were 0.9729/3.3695,0.9101/2.0132 and 0.9807/4.4006 for spectral analysis of the chloropyll a, chloropyll b and carotenoid content per unit volume; When using reflectional hyperspectral imaging technology, the best coefficients of determination of prediction (Rpre/RPD) values of the models were 0.9650/3.4296,0.9008/2.0759 and 0.9813/5.5144 for spectral analysis of the chloropyll a, chloropyll b and carotenoid content per unit volume. In these 3 methods of spectroscopy, the the coefficients of determination of prediction (Rpre) values of the models was only decreased by 0.05%, but RPD averagely increased by 18.68%, in the meanwhile, the mean number of selected effectieve variables was 5.11, equally decreased by 99.55%. The study above showed that these three methods of spectroscopy could effectively and fastly monitor the chloropyll a, b and carotenoid content per volume of Haematococcus pluvialis, in which the reflectional hyperspectral imaging technology got a relatively better result. Moreover, the pigments distribution maps of microalgal liquids were obtained based on the algal hyperspectral images. The results showed that hyperspectral imaging technique is better than RGB images to measure the growing information of microalgae.(3) Feasibility study was carried on that Raman microspectroscopy was used to detect the lipids of microalgae in vivo. Firstly, by Raman microspectroscopy, the mean iodine value (IV) of five speices of microalgae of chlorella pyrenoidosa, Nannochloropsis, Chlorella vulgaris, Chlamydomonas sp. and Haematococcus pluvialis were predicted from 40 sample of each microalgae to respectively 117.69, 121.35,124.50,124.30 and 117.23. In the other way, IV of Chlorella vulgaris detected by GC-MS was compared with the prediction value, which showed only 0.11% differentiation. At the same time, the IVs of the other species of microalgae above were also compared with the measured values refer to published literature and the result showed the feasibility of using Raman microspectroscopy detecting the IV of microalgae by ratio of 1660 cm-1 and 1445 cm-1. In another experiment, chlorella pyrenoidosa cultivated under different nitrogen environments (N starved, N normal and N replete) were detected by Raman microspectroscopy. The mean Raman intensity of lipid peaks of 1066cm-1,1660cm-1 and 1750cm-1 under N starved and N replete condition were compared in third, fifth and seventh day, and as a result the intensity under condition of N starved was higher than N replete. Then, through the ratio of intensity of 1660cm-1 and 1445cm-1, the IV trend of microalgal cells under different days and nitrogen environments were predicted as bubble chart to show a similar trend compared with the Nile Red fluorescence figure. Finally, by chemometrics, different conditions of chlorella pyrenoidosa were discriminated and with characteristic peaks of pigment and lipid, the best coefficients of determination of predition accuracy achiedved 100% by ELM model at the time of fifth day. |
PDF Full Text Request