[Study on the Visualization of the Biomass of Chlorella sp., Isochrysis galbana, and Spirulina sp. Based on Hyperspectral Imaging Technique].

Effective cultivation of the microalgae is the key issue for microalgal bio-energy utilization. In nutrient rich culture conditions, the microalge have a fast growth rate, but they are more susceptible to environmental pollution and influence. So to monitor the the growth process of microalgae is significant during cultivating. Hyperspectral imaging has the advantages of both spectra and image analysis. The spectra contain abundant material quality signal and the image contains abundant spatial information of the material about the chemical distribution. It can achieve the rapid information acquisition and access a large amount of data. In this paper, the authors collected the hyperspectral images of forty-five samples of Chlorella sp., Isochrysis galbana, and Spirulina sp., respectively. The average spectra of the region of interest (ROI) were extracted. After applying successive projection algorithm (SPA), the authors established the multiple linear regression (MLR) model with the spectra and corresponding biomass of 30 samples, 15 samples were used as the prediction set. For Chlorella sp., Isochrysis galbana, and Spirulina sp., the correlation coefficient of prediction (r(pre)) are 0.950, 0.969 and 0.961, the root mean square error of prediction (RMSEP) for 0.010 2, 0.010 7 and 0.007 1, respectively. Finally, the authors used the MLR model to predict biomass for each pixel in the images of prediction set; images displayed in different colors for visualization based on pseudo-color images with the help of a Matlab program. The results show that using hyperspectral imaging technique to predict the biomass of Chlorella sp. and Spirulina sp. were better, but for the Isochrysis galbana visualization needs to be further improved. This research set the basis for rapidly detecting the growth of microalgae and using the microalgae as the bio-energy.

[Identification of Microalgae Species Using Visible/Near Infrared Transmission Spectroscopy].

At present, the identification and classification of the microalgae and its biochemical analysis have become one of the hot spots on marine biology research. Four microalgae species, including Chlorella vulgaris, Chlorella pyrenoidosa, Nannochloropsis oculata, Chlamydomonas reinhardtii, were chosen as the experimental materials. Using an established spectral acquisition system, which consists of a portable USB 4000 spectrometer having transmitting and receiving fiber bundles connected by a fiber optic probe, a halogen light source, and a computer, the Vis/NIR transmission spectral data of 120 different samples of the microalgae with different concentration gradients were collected, and the spectral curves of fourmicroalgae species were pre-processed by different pre-treatment methods (baseline filtering, convolution smoothing, etc. ). Based on the pre-treated effects, SPA was applied to select effective wavelengths (EWs), and the selected EWs were introduced as inputs to develop and compare PLS, Least Square Support Vector Machines (LS-SVM), Extreme Learning Machine (ELM)models, so as to explore the feasibility of using Vis/NIR transmission spectroscopy technology for the rapid identification of four microalgae species in situ. The results showed that: the effect of Savitzky-Golay smoothing was much better than the other pre-treatment methods. Six EWs selected in the spectraby SPA were possibly relevant to the content of carotenoids, chlorophyll in the microalgae. Moreover, the SPA-PLS model obtained better performance than the Full-Spectral-PLS model. The average prediction accuracy of three methods including SPA-LV-SVM, SPA-ELM, and SPA-PLS were 80%, 85% and 65%. The established method in this study may identify four microalgae species effectively, which provides a new way for the identification and classification of the microalgae species. The methodology using Vis/NIR spectroscopy with a portable optic probe would be applicable to a diverse range of microalgae species and proves to be a rapid, real-time, non-destructive, precise method for the physiological and biochemical detection for microalgae.

[Study on the Nondestructive Detection Methods for Dynamica Change of Lipid Content in Chlorella sp.].

Microalgae based biodiesel production requires a large amount of lipid accumulation in the cells, and the accumulation is greatly influenced by the environment. Therefore, it is necessary to find fast and non-destructive methods for lipid change detection. In this paper, Chlorella sp. was adopted as the objective, which was cultured under different light condition consisted of red and blue lights with different proportion. We applied the visible near-infrared spectroscopy (Vis/NIRs) technique to detect the dynamic change of lipid during the microalgae growth processes and utilized hyperspectral imaging technology for visualization of lipid distribution in the suspension. The transmittance and reflectance spectra of microalgae were acquired with Vis/NIRs and hyper-spectroscopy, respectively. In the comparison of the transmittance and reflectance spectra, they showed some different characteristics. Meanwhile it also varied in terms of the number and the area of feature wavelengths obtained by successive projections algorithm (SPA) based on the different spectra. But the established multiple linear regression (MLR) model for lipid content prediction had similar results with rpre of 0.940, RMSEP of 0.003 56 and rpre of 0.932, RMSEP of 0.004 23, respectively. Based on the predictive model, we obtained the spectra and analyzed the lipid dynamic change in microalgae in one life cycle. In the life cycle, the lipid content in Chlorella sp. was relatively stable from the beginning of inoculation to exponential phase, the increase and accumulation of lipid phenomenon occurred in the late exponential phase. Combined with the MLR model and the hypersepctral images, we studied the visualization result of microalgae suspension in the steady phase. The stimulated images showed that the microalgae with higher lipid content appeared gathering. This study compared the difference and the feasibility of the Vis/NIRs and hyperspectral imaging technique in lipid content detection applied in microalgae growing microalgae. The results are meaningful for the fast and non-destructive detection of the growth information of microalgae. It has boththeoretical and practical significance for developing microalgal culture and harvest strategy in practice.

[Application of near-infrared hyperspectral imaging to predicting water content in salmon flesh].

Near-infrared hyperspectral imaging technique was employed in the present study to determine water contents in salmon flesh rapidly and nondestructively. Altogether 90 samples from different positions of salmon fish were collected for hyperspectral image scanning, and mean spectra were extracted from the region of interest (ROI) inside each image. Sixty samples were randomly selected as calibration set, and the remaining 30 samples formed prediction set. The full-spectrum and water contents were correlated using partial least squares regression (PLSR) and least-squares support vector machines (LS-SVM), which were then applied to predict water contents for prediction samples. A novel variable extraction method called random frog was applied to select effective wavelengths (EWs) from the full-spectrum. PLSR and LS-SVM calibration models were established respectively to detect water contents in salmon based on the EWs. Though the performances of EWs-based models were worse than models using full-spectrum, only 12 wavelengths were used to substitute for the original 151 wavelengths, thus models were greatly simplified and more suitable for practical application. For EWs-based PLSR and LS-SVM models, satisfactory results were achieved with correlation coefficient of prediction (Rp) of 0. 92 and 0. 93 respectively, and root mean square error of prediction (RMSEP) of 1. 31% and 1. 18% respectively. The results indicated that near-infrared hyperspectral imaging combined with chemometrics allows accurate prediction of water contents in salmon flesh, providing important reference for the rapid inspection of fish quality.

[Microalgae Species Identification Study with Raman Microspectroscopy Technology].

Identification and classification of microalgae are basis and premise in the study of physiological and biochemical characteristics for microalgae. Microalgae cells mainly consist of five kinds of biological molecules, including proteins, carbonhydrates, lipids, nucleic acids and pigments. These five kinds of biological molecules contents with different ratio in microalgae cells can be utilized to identify microalgae species as a supplement method. This paper investigated the application of Raman microspectroscopy technology in the field of rapid identification on different algae species such as aschlorella sp. and chlamydomonas sp. . Cultivated in the same conditions of culture medium, illumination duration and intensity, these two kinds of species of microalgae cells were immobilized by using agar, and then the samples were placed under 514. 5 nm Raman laser to collect Raman spectra of different growth periods of different species. An approach to remove fluorescence background in Raman spectra called Rolling Circle Filter (RCF) algorithm was adopted to remove the fluorescent background, and then some preprocessing methods were used to offset the baseline and smooth method of Savitzky-Golay was tried to make the spectra curves of total 80 samples smoother. Then 50 samples were randomly extracted from 80 samples for modeling, and the remaining 30 samples for independent validation. This paper adopted different pretreatment methods, and used the partial least squares (PLS) to establish model between the spectral data and the microalgae species, then compared the effects of different pretreatment methods. The results showed that with Raman microspectroscopy technology, the pretreatment method of max-peak ratio standardization was a more effective identification approach which utilizes the different content ratios of pigments of different microalgae species. This method could efficiently eliminate the influence on Raman signal due to different growth stages of microalgae and decomposition of pigments contents of microalgae in vivo. Compared with other traditional classification methods, this method had significant advantages like simpler procedure and shorter testing time, and it can also avoid some subjective measurement errors caused by unskilled operations. If the threshold was set to +/- 0.5, the prediction accuracy can reach 100%, and when the threshold was +/- 0.2, the prediction accuracy reached 86.67%, which proves the proposed new method can be a good approach to identify different algae varieties.

Hyperspectral imaging for mapping of total nitrogen spatial distribution in pepper plant.

Visible/near-infrared (Vis/NIR) hyperspectral imaging was employed to determine the spatial distribution of total nitrogen in pepper plant. Hyperspectral images of samples (leaves, stems, and roots of pepper plants) were acquired and their total nitrogen contents (TNCs) were measured using Dumas combustion method. Mean spectra of all samples were extracted from regions of interest (ROIs) in hyperspectral images. Random frog (RF) algorithm was implemented to select important wavelengths which carried effective information for predicting the TNCs in leaf, stem, root, and whole-plant (leaf-stem-root), respectively. Based on full spectra and the selected important wavelengths, the quantitative relationships between spectral data and the corresponding TNCs in organs (leaf, stem, and root) and whole-plant (leaf-stem-root) were separately developed using partial least-squares regression (PLSR). As a result, the PLSR model built by the important wavelengths for predicting TNCs in whole-plant (leaf-stem-root) offered a promising result of correlation coefficient (R) for prediction (RP = 0.876) and root mean square error (RMSE) for prediction (RMSEP = 0.426%). Finally, the TNC of each pixel within ROI of the sample was estimated to generate the spatial distribution map of TNC in pepper plant. The achievements of the research indicated that hyperspectral imaging is promising and presents a powerful potential to determine nitrogen contents spatial distribution in pepper plant.

[Visualization of the chilling storage time for turbot flesh based on hyperspectral imaging technique].

This study proposed a new method using visible and near infrared (Vis/NIR) hyperspectral imaging for the detection and visualization of the chilling storage time for turbot flesh rapid and nondestructively. A total of 160 fish samples with 8 different storage days were collected for hyperspectral image scanning, and mean spectra were extracted from the region of interest (ROD inside each image. Partial least squares regression (PLSR) was applied as calibration method to correlate the spectral data and storage time for the 120 samples in calibration set. Then the PLSR model was used to predict the storage time for the 40 prediction samples, which achieved accurate results with determination coefficient (R2) of 0.966 2 and root mean square error of prediction (RMSEP) of 0.679 9 d. Finally, the storage time of each pixel in the hyperspectral images for all prediction samples was predicted and displayed in different colors for visualization based on pseudo-color images with the aid of an IDL program. The results indicated that hyperspectral imaging technique combined with chemometrics and image processing allows the determination and visualization of the chilling storage time for fish, displaying fish freshness status and distribution vividly and laying a foundation for the automatic processing of aquatic products.

[Different wavelengths selection methods for identification of early blight on tomato leaves by using hyperspectral imaging technique].

Identification of early blight on tomato leaves by using hyperspectral imaging technique based on different effective wavelengths selection methods (successive projections algorithm, SPA; x-loading weights, x-LW; gram-schmidt orthogonaliza-tion, GSO) was studied in the present paper. Hyperspectral images of seventy healthy and seventy infected tomato leaves were obtained by hyperspectral imaging system across the wavelength range of 380-1023 nm. Reflectance of all pixels in region of interest (ROI) was extracted by ENVI 4. 7 software. Least squares-support vector machine (LS-SVM) model was established based on the full spectral wavelengths. It obtained an excellent result with the highest identification accuracy (100%) in both calibration and prediction sets. Then, EW-LS-SVM and EW-LDA models were established based on the selected wavelengths suggested by SPA, x-LW and GSO, respectively. The results showed that all of the EW-LS-SVM and EW-LDA models performed well with the identification accuracy of 100% in EW-LS-SVM model and 100%, 100% and 97. 83% in EW-LDA model, respectively. Moreover, the number of input wavelengths of SPA-LS-SVM, x-LW-LS-SVM and GSO-LS-SVM models were four (492, 550, 633 and 680 nm), three (631, 719 and 747 nm) and two (533 and 657 nm), respectively. Fewer input variables were beneficial for the development of identification instrument. It demonstrated that it is feasible to identify early blight on tomato leaves by using hyperspectral imaging, and SPA, x-LW and GSO were effective wavelengths selection methods.

[Application of Vis/NIR diffuse reflectance spectroscopy to the rapid detection and identification of tomato fruit via space mutation breeding].

In order to quickly analyze varieties of tomato via space mutation breeding with near infrared spectra, characteristics of the pattern were analyzed by partial least square. The model was built with radial basis function neural network and regarded the compressed data as the input of neural network input vectors. The model regarded the compressed data as the input of neural network input vectors and the training process was speeded up. For one hundred and five fruit samples of CK, M1 and M2 the training model was built. Forty five samples formed the prediction set. The discrimination rate of these two models achieved 95.6% and 97.8%. It offered a new approach to the fast discrimination of varieties of tomato via space mutation breeding.

[Research on discrimination method of tomato via space mutation breeding based on spectroscopy technology].

In order to quickly analyze varieties of tomato via space mutation breeding with near infrared spectra, firstly, principal component analysis was used to analyze the clustering of tomato leaf samples, and then abundant spectral data were compressed by wavelet transform and the model was built with radial basis function neural network, which offered a quantitative analysis of tomato varieties discrimination. The model regarded the compressed data as the input of neural network input vectors and the training process speeded up. One hundred and five leaf samples of CK, M1 and M2 were selected randomly to build the training model, and forty five samples formed the prediction set. The discrimination rate of 97.8% was achieved by this method. It offered a new approach to the fast discrimination of varieties of tomato via space mutation breeding.

[Application of visible/near infrared spectroscopy to discriminating honey brands based on independent component analysis and BP neural network].

Visible/near infrared spectroscopy (Vis/NIRS) appears to be a rapid and convenient non-destructive technique that can measure the quality and compositional attributes of many substances. In the present study, a nondestructive method for the classification of honey brands was developed using Vis/NIRS. The honey brands studied in the research were Feng boshi, Tian ranfeng and Guan shengyuan. The sample set comprised 30 of each brand. Independent component analysis (ICA) was put forwarded to select several optimal wavelengths based on loading weights. Two types of preprocessing (Savitzky-Golay combined with multiplicative scatter correction) were used before the spectral data were analyzed with multivariate calibration methods of artificial neural network (ANN). The absorbance values log (1/T) (T= transmission), corresponding to the wavelengths of 408, 412, 409, 1 000, 468, 462, 408, 400, 997 and 998 nm were chosen as the input data of ANN. The ANN model with three layers was built, and the transfer function of sigmoid was used in each layer. After several trials, the best neural network architecture was obtained with 10 nodes in hidden layers. In the model, the node of input layer, hidden layer, output layer was set to be 9, 10, and 3 respectively, and the goal error was set to be 0. 000 1, the speed of learning was set to be 0.2, the time of training was set to be 1 500. Seventy five samples (25 with each brand) from three brands were selected randomly as calibration set, and the left 15 samples (5 with each brand) were as perdition set. The discrimination rate of 100% was achieved, and the fitting residual was 8. 245 365 x 10(-5). These indicated that the result of honey discrimination was very good based on ICA method, and offer a new approach to the fast discrimination of varieties of honey.

[Research on discrimination method of orange juice variety based on spectroscopy technology].

In order to quickly analyze varieties of orange juice with near infrared spectra, firstly, principal component analysis (PCA) was used to analysze the clustering of orange juice samples, and the characteristic differentia of four orange juice varieties was obtained through qualitative analysis. Then plentiful spectral data were compressed by wavelet transform (WT) and the model was built with radial basis function neural network (RBF-NN), which offered a quantitative analysis of orange juice varieties discrimination. The model regarded the compressed data as the input of RBF-NN input vectors and built a RBF-NN model. Two hundred forty samples from four varieties were selected randomly to build the training model, which in turn was used to predict the varieties of 60 unknown samples. The discrimination rate of 100% was achieved by WT-RBFNN method. It was indicated that wavelet transform combined with RBF-NN is an available method for variety discrimination based on the near infrared reflectance spectroscopy technology. It offered a new approach to the fast discrimination of varieties of orange juice.

[Discrimination among different brands of coffee by using vis-near infrared spectra].

Near infrared spectroscopy technology was used to distinguish three different brands of coffee bought from the supermarket. Two models, PCA-BP and WT-BP, were employed for the analysis and prediction of the samples. The discrimination among the different brands of coffee was based on the combination of the function of data compression in the PCA and WT technology and the ability of learning and prediction of the artificial neural network. In the experiment, 60 samples were used for model calibration and 30 for brand prediction. The result showed that both the PCA-BP and WT-BP models achieved 100% discrimination rate, and the wavelet transforms technology is superior to the principal component analysis both in time-consuming and the capability of data compression. It is indicated that the model set up by the combination of WT technology and BP neural network in the present study is rapid in analysis and precise in pattern discrimination. It can be concluded that a new approach to distinguishing pure coffee was of fered and the result of this experiment established the foundation for the determination of the raw material (coffee bean) of different brands of coffee in the market.

[Discrimination of varieties of apple using near infrared spectra based on principal component analysis and artificial neural network model].

A new method for the discrimination of varieties of apple by means of near infrared spectroscopy (NIRS) was developed. First, principal component analysis (PCA) was used to compress thousands of spectral data into several variables and describe the body of spectra, the analysis suggested that the cumulate reliabilities of PC1 and PC2 (the first two principle components) were more than 98%, and the 2-dimentional plot was drawn with the scores of PC1 and PC2. It appeared to provide the best clustering of the varieties of apple. The loading plot was drawn with PC1 and PC2 through the whole wavelength region. The fingerprint spectra, which were sensitive to the variety of apple, were obtained from the loading plot. The fingerprint spectra were applied as ANN-BP inputs. Seventy five samples from three varieties were selected randomly, then they were used to build discrimination model. This model was used to predict the varieties of 15 unknown samples; the distinguishing rate of 100% was achieved. This model is reliable and practicable. So the present paper could offer a new approach to the fast discrimination of varieties of apple.