The Spectral Image Acquisition And Data Processing For Underwater In-situ Detection

It is very important in marine science to develop the in-situ chemical sensor for theenvironmental monitoring and the exploration. Compared to the electrochemical sensor,the spectroscopic senor has been a promising technique since the advantages ofmulti-element detection, rapid, stand-off etc. However there are two issues to beresolved: a) the wide spectral range with high resolution; b) the data processing; they bothwere studied in this thesis.In order to achieve the time-resolved detection with wide spectral range and highspectra resolution, a detection system was established based on a Folded GratingsSpectrometer and an ICCD camera. In that system the time-resolved images werere-built by using the SDK of the ICCD in the language LabVIEW. In the conversionfrom spectral image to spectrum, a line of image pixels of each sub-grating was calibratedby laser line from OPO tunable laser and the results compared by high-resolutionspectrometer SP-25000i; and the wavelength calibration were realized by polynomialfitting. Through the interception and splicing of nine spectrum ranges, full spectrumbetween300-700nm was obtained. The analysis shows that the pixel wavelengthresolution capability of0.054nm/pixel can be reached with the wavelength accuracy of0.1nm.Calibration software was developed to solve the distortion in grating diffractionimage which is caused by the craftsmanship of making the Folded Gratings Spectrometer.Each sub-grating diffraction image undergoes three steps process of image correction,1-D median filtering and pixel binning, thus the SNR can be increased by6folds on thepremise of the spectral resolution is basically unchanged. The Raman/LIBS jointdetection in solution was carried out simultaneously as the Raman detection of sulfate/nitrate and the LIBS signal of sodium. The results indicate that it is feasible to apply thissystem underwater for in-situ detection.Addressing the application requirements in underwater spectral detection such asreal-time analysis and long-term monitoring etc., the automatic spectrum analysis methodwas developed by using the calibration curve. In this method, the automatic peakseeking is the key for the whole detection. Considering recognition weak peak andoverlapping peaks, the sensibility to noise, the influence of the background and tendency,and the convenience for computer’s automatic processing, the symmetric zero areatransformation method was utilized. To solve the problem that peak parameters arereally difficult to be accurately extracted from overlapping peaks, the preliminaryparameters of overlapping peaks are recognized by symmetrical zero area transformationmethod, and the parameters are fitted and optimized by L-M algorithm, thus theextraction accuracy of peak parameters is improved. In terms of automatic filtering method, the thesis proves the EMD method’s potential in spectral automatic filtering bycomparing the wavelet threshold filtering and the EMD method. In this way, theregression analysis on the spectral data in marine trial data of DOCARS system wasexamined and the ideal result was obtained.After summarizing the completed work, the further research will be focused on thein-situ underwater detection including the automatic wavelength calibration method; theintensity calibration method research in marine water; the in-depth study of automaticspectrum analysis method. The joint detection experiments will be implemented basedon the study of this work.