Research On Rapid Detection Method And Application Of Discrete Three-Dimensional Fluorescence

The structure of plank-tonic algae community is one of the important indicators to evaluate the quality of aquatic ecological environment.When the water environment changes,the structure and content of the algae community will change significantly.At present,the analysis of plank-tonic algae communities is mainly based on "on-site sampling-laboratory analysis".However,the sampling frequency is low,the analysis speed is slow,and the lack of fast online automatic,in-situ profile and remote sensing ground verification equipment cannot meet the requirements of rapid plank-tonic algae communities in a large area of water by online/in-situ monitoring measurement.Due to the advangetages of fast detection speed,high sensitivity,and online monitoring/in-situ measurement applications,the discrete three-dimensional fluorescence spectroscopy technology can provide the new method for on-site rapid measurement of plank-tonic algae communities.Therefore,a rapid detection method of planktonic algae communities in water based on discrete three-dimensional fluorescence spectroscopy technology has been carried out.This technology can realize rapid online/in-situ analysis of the algae community structure in water.Moreover,it can also be used for water ecological environmental pollution monitoring/evaluation and early warning,and provide theoretical data for biodiversity,environmental scientific research,and ground verification of remote sensing data.The research work has important scientific significance and application value.The main results of the thesis were as follows:(1)The method for extracting features of the discrete three-dimensional fluorescence spectra of planktonic algae was studied.The three-dimensional fluorescence spectrum characteristics of common algae in water bodies such as cyanobacteria,green algae,diatoms,dinoflagellate,and cryptophytes were analyzed,and the similarity of the three-dimensional fluorescence spectra of different algae was studied.The results showed that the three-dimensional fluorescence spectra of different species of the same phylum have high similarity.Different phylum algae had obvious differences;the three-dimensional fluorescence spectra of algae based on the elimination of uninformed variables combined with the partial least squares algorithm(UVE-PLS)and the Monte Carlo uninformed variable elimination combined with the partial least squares algorithm(MC-UVE-PLS)had been studied.The characteristic wavelength extraction method obtains 12 characteristic excitation wavelengths and 8 characteristic emission wavelengths to characterize the three-dimensional fluorescence spectra,and realized the reconstruction of the discrete three-dimensional fluorescence spectra of different algae.Among them,the MC-UVE-PLS extraction algorithm had simplified data and higher accuracy.(2)The classification and identification method of the discrete three-dimensional fluorescence spectroscopy of the algae community was studied.The identification method of discrete three-dimensional fluorescence spectrum of algae based on support vector machine and multi-dimensional scale transformation method was also studied.The results showed that the correct rate of the support vector machine method for the identification of cyanobacteria,green algae,diatoms,dinoflagellates,cryptophytes and other pure species of algae were over 70%,the recognition accuracy of mixed algae was 30%;the recognition accuracy rates of the multi-dimensional scaling method for cyanobacteria,green algae,diatoms,dinoflagellate,cryptophytes,and mixed algae are 90%,90%,60%,70%,80%,and 40%,respectively.A comparative study of the algae discrete three-dimensional fluorescence spectrum identification method based on the parallel factor(PARAFAC)algorithm and the self-weighted alternating tri-linear decomposition(SWTATLD)algorithm were carried out.The research results show that the recognition rates of mixed algae based on the PARAFAC algorithm and the SWTATLD algorithm were 86.67%and 93.33%,respectively.So,the SWTATLD algorithm has the higher recognition accuracy.(3)The quantitative analysis method of discrete three-dimensional fluorescence spectroscopy of algae community was studied.The quantitative analysis method of discrete three-dimensional fluorescence spectra of algae communities based on multiple linear regression analysis and convolutional neural network(PlainCNN,TextCNN,and PlainCNN-MT)models were studied.The results showed that the relative standard deviations of the predictions of pure species algae by multiple linear regression analysis were all less than 5%,the relative error between the predicted concentration of mixed algae and the actual concentration was more than 10%.Based on the analysis of the convolutional neural network model,the average accuracy of the convolutional neural network multi-task(PlainCNN-MT)model was 95%.and the average root mean square error was 0.039.the precision was better.(4)The field data comparison(Qiandao Lake and Xiangshan Port)and application analysis(Yellow,Bohai Sea and South China Sea)of the rapid identification and detection method of planktonic algae community discrete three-dimensional fluorescence spectroscopy were carried out.The research method of the thesis was applied to the self-developed rapid online/in-situ measurement instrument for the algae community in the water.The results showed that the measurement results of the research and development instrument up to 0.618,which in good agreement with the laboratory manual microscopic examination results,and it has absolute in the identification of diatoms and dinoflagellates.In the marine survey experiments,the dominant algae species in the Yellow Sea and the South China Sea was dinoflagellate,and the community distribution composition was dinoflagellate>greenalga>diatom>cyanophyte and diatom>diatom>cyanobacteria>greenalgae>cryptophytes,respectively.The long-term online monitoring results of the Muping Coastal Environmental Comprehensive Test Station of the Chinese Academy of Sciences showed that dinoflagellate was the dominant species,followed by cyanobacteria and a small amount of diatoms.The research results of the thesis showed that the rapid detection method of water algae community based on the discrete three-dimensional fluorescence spectroscopy technology can realize the rapid analysis of the water algae community structure and provided advanced technical means for the rapid acquisition of water algae community ecological environment data.