Original Spectral Data Research Articles

Nonlinear spatial decoupling strategy and deformable convnets v2 for hyperspectral image classification

Classification of hyperspectral image (HSI) based on deep learning has attracted extensive attention. However, there is a strong spatial nonlinear coupling phenomenon among data in HSI, which leads to low spatial separability of different types of data and affects classification accuracy. In order to solve this problem, this paper proposes a HSI classification method based on nonlinear spatial decoupling strategy and deformable convnets v2 (DCNv2). Specifically, this paper introduces the classic color names (CN) algorithm with strong nonlinear projection ability in the field of target tracking. Through the projection matrix of CN algorithm, the spatial constraint features of the original spectral data extracted by bilinear local binary pattern and the original spectral data are projected to realize the spatial decoupling of features and improve the spatial separability of data. Then, the DCNv2, which has strong adaptability of geometric transformation, is used to extract the deep features of the decoupled data, so as to achieve effective classification. The experiment was carried out on four standard hyperspectral data sets and the overall accuracy of the proposed method on these data sets reached 99.73%, 99.93%, 99.98% and 99.99%, respectively. Compared with several classical classification methods, it has better classification effect.

Machine learning modeling and prediction of peanut protein content based on spectral images and stoichiometry

For rapid nondestructive detection of peanut protein content, an experimental method combining hyperspectral imaging technology and spectrophotometry was proposed. For data redundancy and noise analysis, ten algorithms were selected for feature extraction, and revealed that the optimal characteristic band of protein content was between 400 and 550 nm. According to the results, the median filtering algorithm (MF) was used to preprocess original spectral data, the XGBoost algorithm was used to extract the top 30 feature bands, the Ridge algorithm was used to construct the protein content prediction model, and the protein content physicochemical data were measured by spectrophotometry. The optimal model was MF-XGBoost-Ridge, with hyperparameter α tuning by Optuna algorithm, with RMSE = 0.009, and a correlation R = 0.886 with a fitting time of only 0.02 s. Compared with the traditional machine learning algorithm models, the prediction accuracy of this study was high and the fitting time was short.

Camouflage Target Recognition Based on Dimension Reduction Analysis of Hyperspectral Image Regions

Hyperspectral reconnaissance technology can realize three-dimensional reconnaissance by using target space and spectral information, which effectively improves the efficiency of battlefield reconnaissance. However, in order to obscure what is true and what is false to confuse the enemy, camouflage technology is also developing. Hiding the target in the background environment and setting false targets have become common camouflage procedures on the battlefield. The camouflaged target has very similar spatial and spectral characteristics to the real target, so the method of identifying the camouflaged target according to the similarity threshold of the original spectral data is no longer reliable. In order to solve the problem of high spectral similarity and low discrimination between a camouflaged target and a real target in a hyperspectral image, a joint processing method of spatial spectrum information is adopted in this paper. Firstly, the hyperspectral image is preprocessed, and then the target area to be measured is determined. Finally, the dimensions of the determined sensitive small area are reduced. Experiments show that this processing method can effectively reduce the spectral similarity of true and false targets, increase the spectral difference of true and false targets and improve the ability to identify true and false targets based on hyperspectral images.

Research on non-destructive testing of hotpot oil quality by fluorescence hyperspectral technology combined with machine learning

Eating repeatedly used hotpot oil will cause serious harm to human health. In order to realize rapid non-destructive testing of hotpot oil quality, a modeling experiment method of fluorescence hyperspectral technology combined with machine learning algorithm was proposed. Five preprocessing algorithms were used to preprocess the original spectral data, which realized data denoising and reduces the influence of baseline drift and tilt. The feature bands extracted from the spectral data showed that the best feature bands for the two-classification model and the six-classification model were concentrated between 469 and 962 nm and 534–809 nm, respectively. Using the PCA algorithm to visualize the spectral data, the results showed the distribution of the six types of samples intuitively, and indicated that the data could be classified. Based on the modeling analysis of the feature bands, the results showed that the best two-classification models and the best six-classification models were MF-RF-RF and MF-XGBoost-LGB models, respectively, and the classification accuracy reached 100 %. Compared with the traditional model, the error was greatly reduced, and the calculation time was also saved. This study confirmed that fluorescence hyperspectral technology combined with machine learning algorithm could effectively realize the detection of reused hotpot oil.

Hyperspectral prediction of mangrove leaf stoichiometries in different restoration areas based on machine learning models

Mangroves play an extremely important role in purifying the atmosphere and responding to global temperature changes. The analysis of chemical elements (carbon, nitrogen, phosphorus, etc.) in mangroves is an effective way to investigate physiological activities, such as vegetation growth, development, and material metabolism. Therefore, the monitoring of mangrove stoichiometry is extremely important for mangrove restoration. Here, two mangrove species, Kandelia candel (KC) and Aegiceras corniculatum (AC), were studied in the Quanzhou Bay Estuary Wetland Nature Reserve. Two machine learning models [random forest (RF) and back propagation neural network (BPNN)] and partial least squares (PLS) were established with the original spectral data as independent variables, and the optimal model was selected by comparing the simple cross-validation VEcv, the ratio of performance to deviation, and the root mean square error (RMSE). The results showed that: (1) the contents of total phosphorous and total nitrogen decreased gradually and the content of total carbon of mangroves increased gradually with an increase in age at restoration; (2) hyperspectral modeling can invert the ecological stoichiometries of KC and AC, and it can be used to effectively monitor the growth status of the species studied; and (3) model performance ranking, PLS > RF > BPNN, where PLS (VEcv ≥ 0.60 and RMSE < 40) was significantly better than the other two models, and BPNN was the least effective and not suitable for hyperspectral inversion modeling of KC and AC ecological stoichiometries. This study provides a methodological basis for long-term and large-scale dynamic monitoring of mangrove ecological stoichiometries and mangrove restoration quality based on hyperspectral data.

Prediction of Maize Seed Vigor Based on First-Order Difference Characteristics of Hyperspectral Data

The identification of seed vigor is of great significance to improve the seed germination rate, increase crop yield, and ensure product quality. In this study, based on a hyperspectral data acquisition system and an improved feature extraction algorithm, an identification model of the germination characteristics for corn seeds was constructed. In this research, hyperspectral data acquisition and the standard corn seed germination test for Zhengdan 958 were carried out. By integrating the hyperspectral data in the spectral range of 386.7–1016.7 nm and the first derivative information of the spectral data, the root length prediction for corn seeds was successfully completed. The data regression model and prediction relationship between the spectral characteristics and seedling root length were established by principal component regression, partial least squares, and support vector regression. The first derivative information of the hyperspectral data was obtained by comparing the prediction model results with the original spectral data, which was preprocessed by Savitzky–Golay smoothing, multiplicative scatter correction, standard normal variate, and curve fitting. The results showed that the prediction model based on the first-order differential spectral data showed better performance than the one based on the spectral data obtained by other processing algorithms. By comparing the prediction results using different data characteristics and regression models, it was found that the hyperspectral method can effectively predict the root length of the seed, with the coefficient of determination reaching 0.8319.

Quantitative Inversion Model Design of Mine Water Characteristic Ions Based on Hyperspectral Technology

In view of the problems of low measurement accuracy and repeated calibration during the use of coal mine water quality analysis, the hyperspectral reflection noncontact measurement technology was proposed to solve the existing problems. KCl, NaCl, pH, NaHCO3, and CaCl2were used to indicate the characteristic ion information of Na+, K+, Ca2+, Cl−, HCO3−, and pH mine water in the laboratory, and 2220 spectral data were obtained by spectral determination. Savitzky–Golay convolution smoothing was used to smooth and denoise the original spectral data of each ion, and the relationship between the spectrum and the concentration of each reagent was obvious after smoothing and denoising pretreatment. The principal component regression method was used to build the inversion model of each ion content, and through the modeling study, the prediction set of KCl was found: the coefficient R∧2 reaches 0.907, RPD is up to 2.7; the prediction set of NaCl was found: the coefficient R∧2 reaches 0.957, RPD is up to 3.1; the PH prediction set was found: the coefficient R∧2 reaches 0.785, RPD is up to 2.1; the prediction set of NaHCO3 was found: the coefficient R∧2 reaches 0.137, RPD is up to 1.2; the prediction set of CaCl2 was found: the coefficient R∧2 reaches 0.622, and RPD is up to 1.7. The results show that the hyperspectral method can play a better role in the extraction of K+, Cl−, Na+, Ca2+, and pH. It is difficult to extract HCO3− ions.

Soluble Solids Content Binary Classification of Miyagawa Satsuma in Chongming Island Based on Near Infrared Spectroscopy.

Citrus is one of the most important fruits in China. Miyagawa Satsuma, one kind of citrus, is a nutritious agricultural product with regional characteristics of Chongming Island. Near-infrared Spectroscopy (NIR) is a proper method for studying the quality of fruits, because it is low-cost, efficient, non-destructive, and repeatable. Therefore, the NIR technique is used to detect citrus's soluble solid content (SSC) in this study. After obtaining the original spectral data, the first 70% of them are divided into the training set and 30% into the test set. Then, the Random Frog algorithm is chosen to select characteristic wavelengths, which reduces the dimension of the data and the complexity of the model, and accordingly makes the generalization of the classification model better. After comparing the performance of various classifiers (AdaBoost, KNN, LS-SVM, and Bayes) under different characteristic wavelength numbers, the AdaBoost classifier outperforms using 275 characteristic wavelengths for modeling eventually. The accuracy, precision, recall, and F1-score are 78.3%, 80.5%, 78.3%, and 0.780, respectively and the ROC (Receiver Operating Characteristic Curve, ROC curve) is close to the upper left corner, suggesting that the classification model is acceptable. The results demonstrate that it is feasible to use the NIR technique to estimate whether the citrus is sweet or not. Furthermore, it is beneficial for us to apply the obtained models for identifying the quality of citrus correctly. For fruit traders, the model helps them to determine the growth cycle of citrus more scientifically, improve the level of citrus cultivation and management and the final fruit quality, and thus increase the economic income of fruit traders.

The Classification of Rice Blast Resistant Seed Based on Ranman Spectroscopy and SVM.

Rice blast is a serious threat to rice yield. Breeding disease-resistant varieties is one of the most economical and effective ways to prevent damage from rice blast. The traditional identification of resistant rice seeds has some shortcoming, such as long possession time, high cost and complex operation. The purpose of this study was to develop an optimal prediction model for determining resistant rice seeds using Ranman spectroscopy. First, the support vector machine (SVM), BP neural network (BP) and probabilistic neural network (PNN) models were initially established on the original spectral data. Second, due to the recognition accuracy of the Raw-SVM model, the running time was fast. The support vector machine model was selected for optimization, and four improved support vector machine models (ABC-SVM (artificial bee colony algorithm, ABC), IABC-SVM (improving the artificial bee colony algorithm, IABC), GSA-SVM (gravity search algorithm, GSA) and GWO-SVM (gray wolf algorithm, GWO)) were used to identify resistant rice seeds. The difference in modeling accuracy and running time between the improved support vector machine model established in feature wavelengths and full wavelengths (200–3202 cm−1) was compared. Finally, five spectral preproccessing algorithms, Savitzky–Golay 1-Der (SGD), Savitzky–Golay Smoothing (SGS), baseline (Base), multivariate scatter correction (MSC) and standard normal variable (SNV), were used to preprocess the original spectra. The random forest algorithm (RF) was used to extract the characteristic wavelengths. After different spectral preproccessing algorithms and the RF feature extraction, the improved support vector machine models were established. The results show that the recognition accuracy of the optimal IABC-SVM model based on the original data was 71%. Among the five spectral preproccessing algorithms, the SNV algorithm’s accuracy was the best. The accuracy of the test set in the IABC-SVM model was 100%, and the running time was 13 s. After SNV algorithms and the RF feature extraction, the classification accuracy of the IABC-SVM model did not decrease, and the running time was shortened to 9 s. This demonstrates the feasibility and effectiveness of IABC in SVM parameter optimization, with higher prediction accuracy and better stability. Therefore, the improved support vector machine model based on Ranman spectroscopy can be applied to the fast and non-destructive identification of resistant rice seeds.

Improved Kernel Entropy Composition Analysis method for transgenic cotton seeds recognition based on Terahertz Spectroscopy

Terahertz (THz) Spectroscopy, featured by low energy, instantaneity and spectral fingerprint, is promising in material identification. Kernel Entropy Composition Analysis (KECA), different from the commonly used THz time-domain spectroscopy analysis methods of PCA and Kernel Principal Component Analysis (KPCA), focuses on decomposing eigenvalues and eigenvectors of a kernel matrix where its original spectral data are projected into a high-dimensional feature space, selecting eigenvectors as projection vectors which contribute most to Rayleigh entropy of original data and getting a new database. However, the parameters selection of kernel function in traditional KECA method have a significant effect on the accuracy of final analysis results. Aimed at this problem, this paper has proposed a new approach called improved KECA method for transgenic cotton seeds recognition based on Terahertz Spectroscopy. The improved KECA takes as the criterion function difference between intra-class and inter-class dispersion based on angular structure for Kernel parameter optimization, and selects kernel parameter with the maximum value of the criterion function as the best option. Then, the clustering method based on angular structure distance is used for the identification of different substance species. In order to test whether the proposed method is effective or not, this paper has applied the THz time-domain spectroscopy technology to detect the three transgenic cotton seeds of Xinluzhong6, Xinqiu107 and Yingmian8 respectively. And then the absorbance spectrum data of these three cotton seeds will be subject to clustering analysis through the proposed improved KECA method.

Discrimination of several robusta organic coffees from Sumatra Indonesia with different origins using UV spectroscopy and principal component analysis

Comparing to non-organic coffee, organically grown coffee has been regarded to be healthier at a more expensive price. Recently, in Sumatra, Robusta organic coffee is becoming more popular. However, it is difficult to assess the origin of organic ground roasted coffee resulting in frequent mislabeling incidence in the market. In this research, an effort to discriminate Robusta organic coffee from different origins was evaluated using ultraviolet (UV) spectroscopy and principal component analysis (PCA). A total of fifty samples of ground-roasted organic coffee from two origins in Sumatra Indonesia (Lampung and Bengkulu) was used. The samples were extracted using hot distilled water, filtered, and diluted with distilled water (1:70 volume/volume). UV spectra were measured in the range of 220-400 nm with 1 nm of the resolution using a benchtop UV-Vis spectrometer. Principal component analysis (PCA) was applied for original and preprocessed spectral data in the range of 220-400 nm. Using the first and the second principal components (PC1 and PC2) using preprocessed spectra resulted in a distinct cluster between Lampung and Bengkulu organic coffee samples. This result showed the effectiveness of PCA in differentiating between the organic Robusta coffee from different origins (Lampung and Bengkulu).

A study on water quality parameters estimation for urban rivers based on ground hyperspectral remote sensing technology.

The purpose of this research is to seek a better inversion algorithm. And on this basis, it explores the feasibility of using hyperspectral monitoring technology instead of laboratory physical and chemical index test and evaluates the prediction effect of inversion model on water quality change. So as to be more convenient, more economical and extensive monitoring methods for water quality monitoring of urban internal river are provided. This paper takes the water samples collected in Fuyang River in downtown Handan as the research object and obtains original spectral data of the samples by the ASD FieldSpec 4 field hyperspectral spectrometer. After the smoothing filter pretreatment by the Savitzky-Golay (SG) method and specified mathematical transformations, the modeling spectral indicators of various water quality parameters are selected and determined by calculating the maximum mean of absolute values for correlation coefficients of various spectral indicators and measured values in the wavelength range from 400 to 950nm. By introducing partial least squares (PLS), random forest (RF), and Lasso (least absolute shrinkage and selection operator), six water quality parameter fitting models were constructed including turbidity (Turb), suspended substance (SS), chemical oxygen demand (COD), NH4-N, total nitrogen (TN), and total phosphorus (TP), which are also testified and evaluated through hyperspectral data. The results show that different spectral transformation methods highlight different information inversion effects. The first derivative of reciprocal logarithm of spectral data after SG smoothing has a good modeling effect on four water quality parameters including Turb, COD, NH4-N, and TP; and the first derivative of smoothed spectral data has a good modeling effect on both water quality parameters of SS and TN. Among the three models, the PLS model has a good prediction effect, with the [Formula: see text] for COD, TN, and TP ranging from 0.74 to 0.80, while that for Turb and SS shows relatively poorer prediction effect, followed by even worse effect on HN4-H. Both machine learning algorithms of RF and Lasso have respectively obtained the best prediction models for different water quality parameters. The Lasso model has a [Formula: see text] value above 0.8 for water body organic pollutants COD, TN, and TP, and the decrease value for [Formula: see text] and [Formula: see text] is below 0.1, which indicates that the model has high prediction accuracy and strong generalization ability, but the results of SS and NH4-N do not meet the expected accuracy. In the inversion model of RF for COD, [Formula: see text] is higher than [Formula: see text], which shows excellent performance, and has certain prediction ability for SS and NH4-N. The RF model and Lasso model complement each other effectively in applicability and prediction accuracy. Compared with the traditional regression model PLS, machine learning has obvious overall advantages, making it more suitable for classified inversion prediction of urban river water quality parameters.

Portable NIR spectroscopy and PLS based variable selection for adulteration detection in quinoa flour

Quinoa flour is prone to economically motivated adulteration due to its high nutritional value and growing demand worldwide. In this study, portable near-infrared spectroscopy (NIRS) combined with multivariate analysis was used to detect adulteration in quinoa flour. The initial investigation was carried out using principal component analysis (PCA) and uniform manifold approximation and projection (UMAP) to identify the most potential adulterate in quinoa flour. Quinoa flour samples were then adulterated with identified adulterant in the range of 0–51% (w/w). The partial least squares regression (PLSR) model built with original spectral data had the best performance for detecting the level of adulteration, with a coefficient of determination (Rp2) of 0.94, root-mean-square error of prediction (RMSEP) of 3.04%, ratio of prediction to deviation (RPD) of 4.04, and range error ratio (RER) of 11.84. The variable importance in projection (VIP) resulting from the PLSR model was used to select 13 informative spectral bands. The new PLSR model led to an R2p of 0.98, RMSEP of 1.60%, RPD of 7.71, and RER of 22.56. The results demonstrated the potential of portable NIRS as a rapid, low cost and non-destructive analytical tool for adulteration detection in quinoa flour.

At-line quality assurance of deep-fried instant noodles using pilot scale visible-NIR spectroscopy combined with deep-learning algorithms

Deep-fried instant noodles produced in a pilot scale facility were monitored for quality parameters using at-line visible-NIR spectroscopy (380–1650 nm) combined with deep-learning algorithms. To build robust calibrations, a wider range of quality parameters for instant noodles viz., moisture (1.6–11.04%), crude protein (8.34–14.39%), total fat (12.38–26.68%), and total ash (1.18–3.15%) were selected. The original raw spectral data was subjected to different pretreatments (standard normal variate (SNV), detrending (DT), first derivative (FD), multiplicative scattering correction (MSC), among others) before being optimized for wavelength selection feature algorithm by competitive adaptive reweighed sampling (CARS). The calibration models were established using deep-learning algorithms based on conventional partial least squares regression (PLSR) and support vector machine regression (SVMR). In general, the SVMR modeling gave an optimum prediction statistics (pretreatment method, coefficient of determination ( R 2 ), root mean square error of prediction (RMSEP), ratio of prediction to deviation (RPD)) for moisture (CARS–SNV–DT, 0.98, 0.32, 6.7), crude protein (CARS–FD, 0.98, 0.18, 8.15), and total fat (CARS–MSC, 0.99, 0.39, 10.15) whereas partial least squares regression (PLSR) gave for total ash (CARS–raw, 0.94, 0.12, 4.34). In particularly for noodle manufacturers, the unification of visible-NIR spectroscopy and deep-learning algorithm is a promising to realize sustainability in quality assurance and control. • Vis-NIR and deep-learning algorithm applied to quality monitoring of noodles. • Calibration models were optimized by CARS feature selection algorithm. • CARS-SVMR gave best predictions for moisture, crude protein, and total fat. • SVMR algorithm successfully used for at-line vis-NIR based quality monitoring.

Application of hyperspectral imaging technology in the rapid identification of microplastics in farmland soil

Microplastics (MPs) are emerging environmental pollutants and their accumulation in the soil can adversely affect the soil biota. This study aims to employ hyperspectral imaging technology for the rapid screening and classification of MPs in farmland soil. In this study, a total of 600 hyperspectral data are collected from 180 sets of farmland soil samples with a hyperspectral imager in the wavelength range of 369– 988 nm. To begin, the hyperspectral data are preprocessed by the Savitzky-Golay (S-G) smoothing filter and mean normalization. Second, principal component analysis (PCA) is used to minimize the dimensions of the hyperspectral data and hence the amount of data, making the subsequent model easier to construct. The cumulative contribution rate of the first three principal components is reached 98.37%, including the main information of the original spectral data. Finally, three models including decision tree (DT), support vector machine (SVM), and convolutional neural network (CNN) are established, all of which can achieve well classification effects on three MP polymers including polyethylene (PE), polypropylene (PP), and polyvinyl chloride (PVC) in farmland soil. By comparing the recognition accuracy of the three models, the classification accuracy of DT and SVM is 87.9% and 85.6%, respectively. The CNN model based on the S-G smoothing filter obtains the best prediction effect, the classification accuracy reaches 92.6%, exhibiting obvious advantages in classification effect. Altogether, these results show that the proposed hyperspectral imaging technique identifies the soil MPs rapidly and nondestructively, and provides an effective automated method for the detection of polymers, requiring only rapid and simple sample preparation.

Case Study on the Fitting Method of Typical Objects

This study proposes different fitting methods for different types of targets in the 400–900 nm wavelength range, based on convex optimization algorithms, to achieve the effect of high-precision spectral reconstruction for small space-borne spectrometers. This article first expounds on the mathematical model in the imaging process of the small spectrometer and discretizes it into an AX=B matrix equation. Second, the design basis of the filter transmittance curve is explained. Furthermore, a convex optimization algorithm is used, based on 50 filters, and appropriate constraints are added to solve the target spectrum. First, in terms of spectrum fitting, six different ground object spectra are selected, and Gaussian fitting, polynomial fitting, and Fourier fitting are used to fit the original data and analyze the best fit of each target spectrum. Then the transmittance curve of the filter is equally divided, and the corresponding AX=B discrete equation set is obtained for the specific object target, and a random error of 1% is applied to the equation set to obtain the discrete spectral value. The fitting is performed for each case to determine the best fitting method with errors. Subsequently, the transmittance curve of the filter with the detector characteristics is equally divided, and the corresponding AX=B discrete equation set is obtained for the specific object target. A random error of 1% is applied to the equation set to obtain the error. After the discrete spectral values are obtained, the fitting is performed again, and the best fitting method is determined. In order to evaluate the fitting accuracy of the original spectral data and the reconstruction accuracy of the calculated discrete spectrum, the three evaluation indicators MSE, ARE, and RE are used for evaluation. To measure the stability and accuracy of the spectral reconstruction of the fitting method more accurately, it is necessary to perform 500 cycles of calculations to determine the corresponding MSE value and further analyze the influence of the fitting method on the reconstruction accuracy. The results show that different fitting methods should be adopted for different ground targets under the error conditions. The three indicators, MSE, ARE, and RE, have reached high accuracy and strong stability. The effect of high-precision reconstruction of the target spectrum is achieved. This article provides new ideas for related scholars engaged in hyperspectral reconstruction work and promotes the development of hyperspectral technology.

Detection of heavy metal lead in lettuce leaves based on fluorescence hyperspectral technology combined with deep learning algorithm

The feasibility analysis of fluorescence hyperspectral imaging technology was studied for the detection of lead content in lettuce leaves. Further, Monte Carlo optimized wavelet transform stacked auto-encoders (WT-MC-SAE) was proposed for dimensionality reduction and depth feature extraction of fluorescence spectral data. The fluorescence hyperspectral images of 2800 lettuce leaf samples were selected and the whole lettuce leaf was used as the region of interest (ROI) to extract the fluorescence spectrum. Five different pre-processing algorithms were used to pre-process the original ROI spectral data including standard normalized variable (SNV), first derivative (1st Der), second derivative (2ndDer), third derivative (3rd Der) and fourth derivative (4th Der). Moreover, wavelet transform stacked auto-encoders (WT-SAE) and WT-MC-SAE were used for data dimensionality reduction, and support vector machine regression (SVR) was used for modeling analysis. Among them, 4th Der tends to be the most useful fluorescence spectral data for Pb content detection at 0.067 ∼ 1.400 mg/kg in lettuce leaves, with Rc2 of 0.9802, RMSEC of 0.02321 mg/kg, Rp2 of 0.9467, RMSEP of 0.04017 mg/kg and RPD of 3.273, and model scale (the number of nodes in the input layer, hidden layer and output layer) was 407-314-286-121-76 under the fifth level of wavelet decomposition. Further studies showed that WT-MC-SAE realizes the depth feature extraction of the fluorescence spectrum, and it is of great significance to use fluorescence hyperspectral imaging to realize the quantitative detection of lead in lettuce leaves.

Salinity Monitoring at Saline Sites with Visible–Near-Infrared Spectral Data

To address the global phenomenon of the salinisation of large land areas, a quantitative inversion model of the salinity of saline soils and soil visible–near-infrared (NIR) spectral data was developed by considering saline soils in Zhenlai County, Jilin Province, China as the research object. The original spectral data were first subjected to Savitzky–Golay (SG) smoothing, multiplicative scattering correction (MSC) pre-processing, and a combined transformation technique. The pre-processed spectral data were then analysed to construct the difference index (DI), ratio index (RI), and normalised difference index (NDI), and the Spearman rank correlation coefficient (r) between these three spectral indices and the salt content in the samples was calculated, while a combined spectral index (r &gt; 0.8) was eventually selected as a sensitive spectral index. Finally, a quantitative inversion model for the salinity of saline soils was developed, and the model’s accuracy was evaluated based on partial least squares regression (PLSR), the random forest (RF) algorithm, and the radial basis function (RBF) neural network algorithm. The results indicated that the inversion of soil salt content using the selected combination of spectral indices based on the RBF neural network algorithm was the most effective, with the prediction model yielding an R2 value of 0.950, a root mean square error (RMSE) of 1.014, and a relative percentage deviation (RPD) of 4.479, which suggested a good prediction effect.

Study on the identification of resistance of rice blast based on near infrared spectroscopy

Rice Blast is the most devastating rice disease which poses a serious threat to the safe production of rice. The most effective way to prevent rice blast is to cultivate the rice varieties that have resistance to the disease, however, traditional resistance testing requires professional personnel, a tedious process, long determination time and high cost. In order to quickly identify different resistant rice seeds which are difficult to distinguish with the naked eye, a rapid non-destructive identification method based on Near-Infrared Spectroscopy (NIRS) was proposed. Four different types of resistant rice seeds (high resistance, high susceptibility, susceptibility and resistance) came from in HeiLongjiang province of China were selected as the research objects. A total of 240 spectral data (60 from each variety) were scanned by the NIR spectrometer. The BP neural network (BP), Support Vector Machines (SVM), Probabilistic Neural Network (PNN) models were established based on the original spectral data in the full-spectrum (11520–4000 cm−1). Among all, Raw-BP has the best identification accuracy which reaches 100% with an iteration time of 869 s. After extracting the feature wavelengths by successive projections algorithm (SPA) on the spectral data, Raw-SPA-BP, Raw-SPA-SVM and Raw-SPA-PNN models were established. The accuracy of these three models didn’t improve. But the iteration time of the SPA-BP model was shortened to 791 s. Another group of BP, SVM, and PNN models were established after using different spectral pretreatment methods and the SPA feature extraction. After Multivariate Scatter Correction (MSC), the accuracy of the MSC-SPA-BP model was still 100% and the iteration time was shortened to 840 s, which is 1/30 of the time at which the original data model was formed. The accuracy of the MSC-SPA-PNN model increased from 60% to 90% and the accuracy of the MSC-SPA-SVM model increased from 60% to 85%. Based on the comparison analysis of the models mentioned above, a best neural network identification model of the MSC-SPA-BP with 513 inputs, 8 hidden layers and 4 outputs was established. Its classification accuracy reached 100% with an iteration time of 29 s, indicating that the MSC-SPA-BP model can completely achieve identification of four different resistant rice seeds. Therefore, the proposed method of the BP neural network identification model based on NIRS can be fully applied to the non-destructive rapid identification of rice seeds. Meanwhile, it provides a reference for the rapid identification of other crop seeds.

Research on Infrared Spectral Quantitative Analysis of Hydrocarbon Gases Based on Adaptive Boosting Classifier and PLS

The traditional gas logging using hydrogen flame gas chromatography is gradually difficult to meet the requirements of the current complex oil and gas reservoir exploration because of its long detection cycle, many additional equipment, complex use, maintenance process and other defects. Infrared spectroscopy has obvious advantages in detection speed and convenience. But when it comes to homologues with similar molecular structure, it is difficult to quantitatively analyze the component concentration because of the serious overlap of characteristic absorption peaks. To solve this problem, this paper uses the self-developed experimental platform to carry out a large number of infrared spectral experimental studies on the single substance and six component mixtures. Based on a large number of experimental data, a method combining the AdaBoost classifier with optimization window partial least squares (PLS) modeling is proposed. Through the PLS local modeling for the spectra of different components and different concentration ranges, the boosting algorithm is used to classify the original spectral data, and then the component concentration is calculated. This method solves the problems of difficult identification and inaccurate quantitative analysis of alkane mixed gas components in traditional methods, and effectively improves the detection accuracy of infrared spectrum quantitative analysis of mixed components, which provides the core theoretical support for infrared spectroscopy to replace hydrogen flame gas chromatography which is widely used in gas logging.