Successive Projections Algorithm Research Articles

A detection method of Saposhnikovia divaricata seed vigor based on near-infrared spectral feature extraction

Saposhnikovia divaricata is an important Chinese herbal medicine, the rapid detection of seed vigor is of great significance and practical value for the breeding selection and cultivation production of Saposhnikovia divaricata varieties. It was difficult to solve the problem of rapid testing technology of Saposhnikovia divaricate seed of the shortcomings of traditional chemical testing methods, such as cumbersome operation, time-consuming, easy to cause seed damage, and so on. In this paper, a rapid method for detecting seed vigor was processed for the Saposhnikovia divaricata based on near-infrared spectral feature extraction. First, 612 sets of near-infrared spectral data were obtained using a Fourier transform near-infrared spectrometer, and randomly divided into a calibration set and prediction set according to the ratio of 3:1. Then, the spectral data were preprocessed by the multivariate scatter correction method (MSC), the successive projection algorithm (SPA) was used to extract the feature wave numbers of spectral data. A total of 5 feature wave numbers were selected from the original 1845 wave numbers, with a streamlining rate of 99.73 %. Finally, a detection method of seed vigor was constructed based on the radial basis function neural network (RBF) with 5 feature wave numbers as input data. A network structure of the model MSC-SPA-RBF is of type 5-48-3, the accuracy of seed vigor detection was 99.0 %, and the detection time was 0.0033 s. The result indicates that the established model is a fast, non-destructive, and accurate method for detecting seed vigor of Saposhnikovia divaricata, which provides a theoretical basis and technical support for the rapid detection of Saposhnikovia divaricata seed vigor.

Classification of Garlic (Allium sativum L.) Crops by Fertilizer Differences Using Ground-Based Hyperspectral Imaging System

In contemporary agriculture, enhancing the efficient production of crops and optimizing resource utilization have become paramount objectives. Garlic growth and quality are influenced by various factors, with fertilizers playing a pivotal role in shaping both aspects. This study aimed to develop classification models for distinguishing garlic fertilizer application differences by employing statistical and machine learning techniques, such as partial least squares (PLS), based on data acquired from a ground-based hyperspectral imaging system in the agricultural sector. The garlic variety chosen for this study was Hongsan, and the fertilizer application plots were segmented into three distinct sections. Data were acquired within the VIS/NIR wavelength range using hyperspectral imaging. Following data acquisition, the standard normal variate (SNV) pre-processing technique was applied to enhance the dataset. To identify the optimal wavelengths, various techniques such as sequential forward selection (SFS), successive projections algorithm (SPA), variable importance in projection (VIP), and interval partial least squares (iPLS) were employed, resulting in the selection of 12 optimal wavelengths. For the fertilizer application difference model, six integrated vegetation indices were chosen for comparison with existing growth indicators. Using the same methodology, the model construction showed accuracies of 90.7% for PLS. Thus, the proposed model suggests that efficient regulation of garlic fertilizer application can be achieved by utilizing statistical and machine learning techniques.

Identification of MDA in seized ecstasy-like samples using atmospheric solids analysis probe mass spectrometry and machine learning

In this study, a total of 64 samples of ecstasy tablets seized during operations conducted within the State of Rio Grande do Sul – Brazil in 2021–2022 were analyzed. The instrumental analysis was performed using the Radian ASAP MS instrument with a single quadrupole mass spectrometer and an Atmospheric Solids Analysis Probe (ASAP) interface. Data preprocessing was conducted using MATLAB to improve data quality for subsequent chemometric analyses. Classification models seeking to identify samples containing MDA (3,4-methylenedioxyamphetamine) were developed using Linear Discriminant Analysis (LDA) with variable selection through the genetic algorithm (GA), stepwise (SW), and successive projections algorithm (SPA). Additionally, Partial Least Squares Discriminant Analysis (PLS-DA) models were explored. The GA-LDA model achieved an accuracy of approximately 95% for both training and testing sets, demonstrating its effectiveness in distinguishing between sample classes. It selected only five variables during model building. The SW-LDA model displayed exceptional performance in the training set, achieving 100% for all metrics. However, its accuracy for the testing set was slightly lower at 84.7%, suggesting potential overfitting due to the inclusion of numerous variables. The SPA-LDA model, while not the best performer in the testing set, selected variables that were more easily attributed to specific ions/fragments, aiding in the interpretation of results. Interpretation of selected variables revealed key m/z values associated with different compounds found in the samples, such as MDMA, cocaine, and amphetamine-related fragments. Notably, the SPA method selected six variables that played a crucial role in distinguishing between sample groups. These findings contribute to the advancement of forensic drug analysis and quality control processes.

Identification and sorting of impurities in tea using spectral vision

Removing impurities from Puer tea is an important step in its processing. However, existing methods are unable to accurately and efficiently sort impurities from Puer tea. This study proposes a novel method based on spectral imaging technology to identify impurities in Puer tea. Hyperspectral images of tea and impurities were acquired in the 400–1000 nm wavelengths. The spectra of each category of samples were extracted from the hyperspectral images and comprehensively analysed and processed, mainly including outlier detection and elimination, spectral preprocessing, and characteristic wavelengths selection. Support vector machine (SVM) models are built based on full-spectrum data and characteristic wavelengths data to achieve pixel-level classification of hyperspectral images. The results showed that the best model based on full spectra achieved an accuracy of 97.8% on the test dataset. The best models based on characteristic wavelengths selected by successive projections algorithm and genetic algorithm achieved accuracies of 95.9% and 94.9%, respectively. Finally, these pixel classification models were applied to the test samples to test the effectiveness of the models in identifying impurities. The results indicate that the three selected models can effectively identify impurities and even perform well on identifying unfamiliar categories of impurities.

Early detection of gray blight in tea leaves and rapid screening of resistance varieties by hyperspectral imaging technology.

Gray blight (GB) is a significant disease of tea leaves, posing a severe threat to both the yield and quality. In this study, the process of leaf infection by a pathogenic isolate of the GB disease (DDZ-6) was simulated. Hyperspectral images of normal leaves, infected leaves without symptoms, and infected leaves with mild and moderate symptoms were collected. Combining convolution neural network (CNN), long short-term memory (LSTM), and support vector machine (SVM) algorithms, the early detection model of GB disease, and the rapid screening model of resistant varieties were established. The generality of this method was verified by collecting datasets under field conditions. The visible red-light band demonstrated a pronounced responsiveness to GB disease, with three sensitive bands identified through rigorous screening processes utilizing uninformative variable elimination (UVE), competitive adaptive reweighted sampling (CARS), and the successive projections algorithm (SPA). The 693, 727, and 766 nm bands emerged as highly sensitive indicators of GB. Under ideal conditions, the CARS-LSTM model excelled in early detection of GB, achieving an accuracy of 92.6%. However, under field conditions, the combination of 693 and 727 nm bands integrated with a CNN provided the most effective early detection model, attaining an accuracy of 87.8%. For screening tea varieties resistant to GB, the SPA-LSTM model excelled, achieving an accuracy of 82.9%. This study provides a core algorithm for a GB disease instrument with detection capabilities, which is of great importance for the early prevention of GB disease in tea plantations. © 2024 Society of Chemical Industry.

Identification of fermented soy sauce and blended soy sauce based on dielectric spectra

A new method based on dielectric properties has been developed to distinguish blended soy sauce from fermented soy sauce. Fifty samples of pure fermented soy sauce and blended soy sauce with different contents of hydrolyzed vegetable protein (HVP) were prepared. The dielectric constant and dielectric loss factor of all soy sauce samples in the 30 MHz–3000 MHz frequency range were measured using an impedance analyzer. The sample set was divided into a correction set and a prediction set using the joint x-y distances (SPXY) algorithm, and the partial least squares (PLS) and support vector machine (SVM) models were adopted to distinguish the different samples. The effects of selecting characteristic variables on model prediction using the full spectra (FS), principal component analysis (PCA), and successive projection algorithm (SPA) were compared. Results indicate that the discriminant effect of the PLS model was better than that of the SVM model overall. The SPA–PLS model had the best predictive performance among the six developmental models. The correlation coefficients of the correction set and prediction set were 0.9205 and 0.9096, respectively. The root mean square error of the calibration set was 1.3699 and that of the prediction set was 1.5950. The study demonstrated that the combination of dielectric spectra and stoichiometry can be utilized to determine whether soy sauce has been adulterated.

Near-infrared spectroscopy and multivariate analysis as effective, fast, and cost-effective methods to discriminate Candida auris from Candida haemulonii.

Candida auris and Candida haemulonii are two emerging opportunistic pathogens that have caused an increase in clinical cases in the recent years worldwide. The differentiation of some Candida species is highly laborious, difficult, costly, and time-consuming depending on the similarity between the species. Thus, this study aimed to develop a new, faster, and less expensive methodology for differentiating between C. auris and C. haemulonii based on near-infrared (NIR) spectroscopy and multivariate analysis. C. auris CBS10913 and C. haemulonii CH02 were separated in 15 plates per species, and three isolated colonies of each plate were selected for Fourier transform near-infrared (FT-NIR) analysis, totaling 90 spectra. Subsequently, principal component analysis (PCA) and variable selection algorithms, including the successive projections algorithm (SPA) and genetic algorithm (GA) coupled with linear discriminant analysis (LDA), were employed to discern distinctive patterns among the samples. The use of PCA, SPA, and GA algorithms associated with LDA achieved 100% sensitivity and specificity for the discriminations. The SPA-LDA and GA-LDA algorithms were essential in selecting the variables (infrared wavelengths) of most importance for the models, which could be attributed to binding of cell wall structures such as polysaccharides, peptides, proteins, or molecules resulting from yeasts' metabolism. These results show the high potential of combined FT-NIR and multivariate analysis techniques for the classification of Candida-like fungi, which can contribute to faster and more effective diagnosis and treatment of patients affected by these microorganisms.

Research on hyperspectral timely monitoring model of green tea processing quality based on PSO-LSSVR

To monitor the processing status and content of the main components of tea leaves during green tea processing timely, a monitoring model for the contents of green tea processing based on hyperspectral imaging is established in this paper. The aim of this model is to improve the quality of finished green tea products by addressing the shortcomings of existing methods that cannot achieve timely, non-destructive, and accurate detection. Initially, hyperspectral and biochemical data were collected across four processing procedures of green tea. Following the exclusion of abnormal samples, preprocessing techniques including Savitzky-Golay, multiple scatter correction, and 2nd derivative were compared and selected. Subsequently, feature wavelengths extracted through correlation analysis, principal component analysis, and successive projections algorithm were utilized to construct least squares support vector regression (LSSVR) prediction models, determining the optimal feature extraction method. Finally, particle swarm optimization (PSO) was employed to refine LSSVR model parameters, resulting in the PSO-LSSVR model. Post-optimization, the determination coefficient (R2) exceeded 0.9, the root means square error remained around 0.005, and the residual prediction deviation surpassed 3. The PSO-LSSVR model exhibited better predictive capabilities compared to the LSSVR model, offering a method reference for timely monitoring of green tea processing inclusions.

Apple firmness detection method based on hyperspectral technology

Firmness is a key indicator of apple quality. Building a predictive model for apple firmness based on hyperspectral technology and regression algorithms can achieve rapid, non-destructive, and high-throughput detection of apple firmness. This paper adopts an Adaptive Window Length Savitzky-Golay Smoothing (AWL-SG smoothing) algorithm based on the Savitzky-Golay Smoothing (SG smoothing) algorithm, which can adaptively adjust the window length according to the change rate of spectral data at different wavelengths. SG smoothing, AWL-SG smoothing, Standard Normal Variate (SNV), and Multiplicative Scatter Correction (MSC) algorithms were used to preprocess the original spectral data, and Partial Least Squares (PLS), Ridge Regression (Ridge), and Kernel Ridge Regression (Kernel Ridge) predictive models were constructed to analyze the impact of different preprocessing methods on model prediction accuracy. The prediction models established with spectral data preprocessed by SG smoothing and AWL-SG smoothing algorithms showed significant improvement in predictive performance on the basis of the original spectral data, among which the AWL-SG smoothing algorithm performed the best. The Ridge model established with spectra data preprocessed by AWL-SG smoothing achieved an R2 of 0.8914 in the test set. Successive Projection Algorithm (SPA), Principal Component Analysis (PCA), and Independent Component Analysis (ICA) dimensionality reduction algorithms were used to reduce the dimensions of the full-band spectral data preprocessed by SG smoothing and AWL-SG smoothing algorithms, and Ridge and Kernel Ridge prediction models were constructed. The results showed that both SPA and PCA algorithms could improve the predictive performance of the models, with the PCA performing the best. The combination of AWL-SG + PCA + Ridge achieved the best predictive effect, with an R2 of 0.9146 in the test set.

Hyperspectral imaging coupled with deep learning model for visualization and detection of early bruises on apples

Early bruise on apples caused by external impacts during the transportation process is commonly difficult to be detected on the apple surface, limiting the application of traditional machine vision methods in determining fruit quality. In recent years, hyperspectral imaging (HSI) has emerged as a promising technology for identifying early bruise of fruits due to its efficient and nondestructive detection. In this study, HSI data in the shortwave infrared range were collected at 2-hour and 6-hour intervals after mechanical damage. The combination of the successive projections algorithm (SPA) and principal component analysis (PCA) was used to select three key feature bands, namely, 1074 nm, 1269 nm and 1441 nm. Pseudo color transformation and band ratio algorithm were then employed to improve the contrast between damaged and healthy apple tissues for image enhancement. The fast and precise YOLOv5 (FP-YOLOv5) model achieved effective identification of apple bruises, with a high recognition rate of 95 % and a fast detection speed at 130 fps. Overall, the proposed framework based on band selection and image enhancement exhibits better performance in the detection of early apple bruises, providing useful insights for HSI combined with a deep learning model in the grading evaluation of fruit quality.

Quantitative prediction of soil chromium content using laboratory-based visible and near-infrared spectroscopy with different ensemble learning models

Soil heavy metal chromium pollution poses significant threats to human health and ecosystems, necessitating accurate quantitative prediction methods for effective monitoring and management. This study aims to develop robust predictive models for soil chromium content in farmland soils of Mojiang Hani Autonomous County, Pu’er City, Yunnan Province, China. These models utilize ensemble learning techniques based on visible and near-infrared spectroscopy. Operations before model building involved partitioning datasets with the Kennard-Stone algorithm to ensure representative training and testing sets. Visible and near-infrared spectroscopy-data preprocessing was performed using Savitzky-Golay smoothing and first-order derivative transformations to enhance signal quality. Bands selection was achieved through the Successive Projections Algorithm (SPA), effectively reducing data dimensionality and collinearity. Six ensemble learning models were constructed and assessed for their predictive performance: Bagging-DTR, Random Forest (RF), Adaboost-DTR, XGBoost-DTR, Stacking-1, and Stacking-2. These models utilized Decision Trees (DTR) and Linear Regression (LR) as base learners. Results demonstrated that ensemble models significantly outperformed individual base learners. Notably, the Stacking-2 model achieved the highest accuracy with an R^2 of 0.954, RMSE of 125.967 mg/kg, and RPD of 4.667. To validate the model’s practical applicability, spatial interpolation of soil Cr content was conducted using the Kriging method based on Stacking-2 model predictions. The spatial distribution maps of measured and predicted values exhibited high congruence, underscoring the model’s effectiveness in accurately mapping Cr distribution across the study area. This study underscores the efficacy of integrating ensemble learning with visible and near-infrared spectroscopy-data preprocessing and SPA for precise soil heavy metal prediction. The findings offer valuable insights and a scientific basis for enhanced soil quality monitoring, environmental risk assessment, and informed agricultural land management and pollution control.

Study on Rapid Non-Destructive Detection Method of Corn Freshness Based on Hyperspectral Imaging Technology.

(1) Background: To achieve the rapid, non-destructive detection of corn freshness and staleness for better use in the storage, processing and utilization of corn. (2) Methods: In this study, three varieties of corn were subjected to accelerated aging treatment to study the trend in fatty acid values of corn. The study focused on the use of hyperspectral imaging technology to collect information from corn samples with different aging levels. Spectral data were preprocessed by a convolutional smoothing derivative method (SG, SG1, SG2), derivative method (D1, D2), multiple scattering correction (MSC), and standard normal transform (SNV); the characteristic wavelengths were extracted by the competitive adaptive reweighting method (CARS) and successive projection algorithm (SPA); a neural network (BP) and random forest (RF) were utilized to establish a prediction model for the quantification of fatty acid values of corn. And, the distribution of fatty acid values was visualized based on fatty acid values under the corresponding optimal prediction model. (3) Results: With the prolongation of the aging time, all three varieties of corn showed an overall increasing trend. The fatty acid value of corn can be used as the most important index for characterizing the degree of aging of corn. SG2-SPA-RF was the quantitative prediction model for optimal fatty acid values of corn. The model extracted 31 wavelengths, only 12.11% of the total number of wavelengths, where the coefficient of determination RP2 of the test set was 0.9655 and the root mean square error (RMSE) was 3.6255. (4) Conclusions: This study can provide a reliable and effective new method for the rapid non-destructive testing of corn freshness.

Estimation of Chlorophyll Content in Apple Leaves Infected with Mosaic Disease by Combining Spectral and Textural Information Using Hyperspectral Images

Leaf chlorophyll content (LCC) is an important indicator of plant nutritional status and can be a guide for plant disease diagnosis. In this study, we took apple leaves infected with mosaic disease as a research object and extracted two types of information on spectral and textural features from hyperspectral images, with a view to realizing non-destructive detection of LCC. First, the collected hyperspectral images were preprocessed and spectral reflectance was extracted in the region of interest. Subsequently, we used the successive projections algorithm (SPA) to select the optimal wavelengths (OWs) and extracted eight basic textural features using the gray-level co-occurrence matrix (GLCM). In addition, composite spectral and textural metrics, including vegetation indices (VIs), normalized difference texture indices (NDTIs), difference texture indices (DTIs), and ratio texture indices (RTIs) were calculated. Third, we applied the maximal information coefficient (MIC) algorithm to select significant VIs and basic textures, as well as the tandem method was used to fuse the spectral and textural features. Finally, we employ support vector regression (SVR), backpropagation neural network (BPNN), and K-nearest neighbors regression (KNNR) methods to explore the efficacy of single and combined feature models for estimating LCC. The results showed that the VIs model (R2 = 0.8532, RMSE = 2.1444, RPD = 2.6179) and the NDTIs model (R2 = 0.7927, RMSE = 2.7453, RPD = 2.2032) achieved the best results among the single feature models for spectra and texture, respectively. However, textural features generally exhibit inferior regression performance compared to spectral features and are unsuitable for standalone applications. Combining textural and spectral information can potentially improve the single feature models. Specifically, when combining NDTIs with VIs as input parameters, three machine learning models outperform the best single feature model. Ultimately, SVR achieves the highest performance among the LCC regression models (R2 = 0.8665, RMSE = 1.8871, RPD = 2.7454). This study reveals that combining textural and spectral information improves the quantitative detection of LCC in apple leaves infected with mosaic disease, leading to higher estimation accuracy.

Detection of dried jujube from fresh jujube with different variety and maturity after hot air drying based on hyperspectral imaging technology

The consistent quality of dried jujube has a significant impact on its economic value and consumer acceptance. Currently, the evaluation of dried jujube quality heavily relies on agricultural experts and advanced instrument analysis, which is time costly and is limited by many uncertainties. This study aims to develop a rapid and accurate method for classifying dried jujube produced from fresh jujube at different stages of maturity using hyperspectral imaging. Dried jujube samples were collected from two varieties (Zhongyang Muzao jujube and Yuanling jujube) in Hebei Province, representing two maturity stages: Kimri and Rutab. Linear partial least squares-discriminant analysis (PLS-DA), K-Nearest Neighbor (KNN) and support vector machine (SVM)algorithms were combined with six pre-processing methods. the pretreatment spectra of Area Naturalization (AN), Baseline and Multiple scattering correction (MSC) were selected for further analysis. Successive projections algorithm (SPA), Competitive adaptive reweighted sampling （CARS）and Uninformative Variables Elimination (UVE)were combined to select characteristic variables. A total of 16 characteristic variables were selected by AN-UVE-SPA-SVM model. This model achieved the highest accuracies of 96.0 % and 93.1 % in the final discriminant accuracy for the calibration and Validation sets, respectively. The hardness, soluble solid content (SSC), and moisture of two categories of dried jujube derived from fresh jujube at various stages of maturity were analyzed. The results showed that dried jujube produced from fresh jujubes at the Rutab stage were more likely to exhibit characteristics of high SSC and low hardness' compared to those produced at the Kimri stage. The partial least squares regression (PLSR) model established based on characteristic wavelengths extracted by AN-UVE-SPA showed the best results in predicting SSC and moisture. This study showcased the feasibility of identifying and classifying dried jujube using hyperspectral imaging technology, laying a solid foundation for further research in the rapid non-destructive testing of dried fruits and vegetables through the application of hyperspectral technology.

Predicting calorific value and ash content of sand shrub using Vis-NIR spectra and various chemometrics

Calorific value (CV) reflects the ability of material flow and energy conversion of plants, which is the key indices of combustion properties for utilization and development of energy plants. However, the commonly used method for CV determination of solid fuels is bomb calorimetry in the laboratory using powder samples, which hinders the capability of rapid and non-destructive prediction for a large-scale samples in a natural environment. Visible and near infrared spectroscopy (Vis-NIR) has been widely proposed as a replace for laboratory determination in properties prediction. However, chemometrics are essential for spectral analysis. Various chemometrics including competitive adaptive reweighted sample (CARS), lifting wavelet transform (LWT), successive projections algorithm (SPA), and convolutional neural networks (CNNs) optimized by whale optimization algorithm (WOA) were employed to optimize models. Additionally, canopy spectra were measured in the field instead of powder samples’ spectra collecting from laboratory. The results demonstrated that CARS-WOA-CNN was the best to predict CV and ash content (AC) with R2 of 0.858 and 0.751, respectively. Compared to raw full spectra, spectral dimension was reduced from 2048 to 93 and 22 for CV and AC, respectively. Overall, this study provided a meaningful strategy for harvest planning and assessing value of biomass in the field.

Protein content prediction of rice grains based on hyperspectral imaging

This study utilized hyperspectral imaging technology combined with mathematical modeling methods to predict the protein content of rice grains. Firstly, the Kjeldahl method was used to determine the protein content of rice grains, and different preprocessing techniques were applied to the spectral information. Then, a prediction model for rice grain protein content was developed by combining the spectral data with the protein content. After performing multiplicative scatter correction (MSC) preprocessing and selecting feature wavelengths based on successive projections algorithm (SPA), the multivariate linear regression (MLR) model showed the best prediction performance, with a calibration set R2C of 0.9393, a validation set R2V of 0.8998, an RMSEV of 0.1725, and an RPD of 3.16. Finally, the quantitative protein content model was mapped pixel by pixel to visualize the distribution of rice protein, providing possibilities for non-destructive protein content detection.

Potential of hyperspectral imaging for nondestructive determination of α-farnesene and conjugated trienol content in ‘Yali’ pear

The sesquiterpene α-farnesene and its corresponding oxidation products, namely conjugated trienols (CTols) is well known to be correlated with the development of superficial scald, a typical physiological disorder after a long term of cold storage in pear fruit. In this work, hyperspectral imaging (HSI) technology was used for nondestructive predicting of α-farnesene and CTols [CT258, CT281 and CT(281-290)] content in ‘Yali’ pear. In order to obtain the best performance of calibration model and simplify the calibration model further, various preprocessing methods together with their combinations and different wavelength selection algorithms, including successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS) and uninformative variable elimination (UVE), were investigated and compared based on linear partial least square regression (PLSR) and nonlinear least square support vector machine (LS-SVM) models, respectively. In conclusion, compared to the PLSR models, the results of LS-SVM models based on original and preprocessing methods performed better for the prediction of α-farnesene and CTols, while the performance of LS-SVM models based on the selected characteristic wavelengths were worse. For α-farnesene, the best result was obtained by LS-SVM model based on MSC-FD pretreatment with the RPD value of 2.6, Rp = 0.925 and RMSEP = 4.387 nmol cm−2. And for CTols, CT281 performed better compared with CT258 and CT(281-290), achieving the result with RPD = 2.4, Rp = 0.913 and RMSEP = 2.734 nmol cm−2 based on LS-SVM model combined with SD pretreatment. The overall results illustrated HSI technology could be used for rapid and nondestructive prediction of α-farnesene and CTols in ‘Yali’ pear, which would be helpful for supporting postharvest decision systems.

Coupling the PROSAIL Model and Machine Learning Approach for Canopy Parameter Estimation of Moso Bamboo Forests from UAV Hyperspectral Data

Parameters such as the leaf area index (LAI), canopy chlorophyll content (CCH), and canopy carotenoid content (CCA) are important indicators for evaluating the ecological functions of forests. Currently, rapidly developing unmanned aerial vehicles (UAV) equipped with hyperspectral technology provide advanced technical means for the real-time dynamic acquisition of regional vegetation canopy parameters. In this study, a hyperspectral sensor mounted on a UAV was used to acquire the data in the study area, and the canopy parameter estimation model of moso bamboo forests (MBF) was developed by combining the PROSAIL radiative transfer model and the machine learning regression algorithm (MLRA), inverted the canopy parameters such as LAI, CCH, and CCA. The method first utilized the extended Fourier amplitude sensitivity test (EFAST) method to optimize the global sensitivity analysis and parameters of the PROSAIL model, and the successive projections algorithm (SPA) was used to screen the characteristic wavebands for the inversion of MBF canopy parameter inversion. Then, the optimized PROSAIL model was used to construct the ‘LAI-CCH-CCA-canopy reflectance’ simulation dataset for the MBF; multilayer perceptron regressor (MLPR), extra tree regressor (ETR), and extreme gradient boosting regressor (XGBR) employed used to construct PROSAIL_MLPR, PROSAIL_ETR, and PROSAIL_XGBR, respectively, as the three hybrid models. Finally, the best hybrid model was selected and used to invert the spatial distribution of the MBF canopy parameters. The following results were obtained: Waveband sensitivity analysis reveals 400–490 and 710–1000 nm as critical for LAI, 540–650 nm for chlorophyll, and 490–540 nm for carotenoids. SPA narrows down the feature bands to 43 for LAI, 19 for CCH, and 9 for CCA. The three constructed hybrid models were able to achieve high-precision inversion of the three parameters of the MBF, the model fitting accuracy of PROSAIL_MLRA reached more than 95%, with lower RMSE values, and the PROSAIL_XGBR model yielded the best fitting results. Our study provides a novel method for the inversion of forest canopy parameters based on UAV hyperspectral data.

Application of Hyperspectral Imaging for Watermelon Seed Classification Using Deep Learning and Scoring Mechanism

Watermelon seeds are a significant source of nutrition in the diet. To assess the potential of near-infrared hyperspectral imaging technology for swift and nondestructive identification of watermelon seed varieties, near-infrared hyperspectral imaging (NIR-HSI) technology was used. The Savitzky–Golay (SG) smoothing algorithm and standard normal variable (SNV) algorithm were combined to preprocess the extracted spectral data. The successive projections algorithm (SPA) was used to reduce the dimensionality of the spectral data. Subsequently, three deep learning models (LeNet, GoogLeNet, and ResNet) were used to classify 10 common watermelon seeds. SPA was used to reduce the dimensionality of hyperspectral data. In terms of full band, the ResNet model achieved a classification accuracy of 86.77% on the test set. By using characteristic bands, the GoogLeNet model achieved a classification accuracy of 83.85% on the test set. The ensemble fusion model based on a scoring mechanism achieved accuracy rates of 99.56%, 90.88%, and 87.97% on the training, validation, and test sets, respectively. The results indicated that the ensemble fusion model based on a scoring mechanism can enhance accuracy. Combining deep learning with NIR-HSI can effectively distinguish different varieties of watermelon seeds.

Rapid and Non-Destructive Determination of the Protein and Fat Contents in Wheat by Hyperspectral Imaging Combined with AdaBoost-SVR Modeling

Wheat protein and fat contents directly affect the flavor and yield of liquor. This study utilized hyperspectral imaging (HSI) technology along with an integrated learning model to rapidly and non-destructively detect protein and fat contents in wheat. The hyperspectral data of the wheat grains was preprocessed using various methods. The competitive adaptive reweighted sampling (CARS) algorithm, the successive projections algorithm (SPA), and the combination of the two algorithms (CARS-SPA) were used to extract the characteristic wavelengths. Three support vector regression (SVR) models were created to forecast wheat protein and fat contents using full-band spectral data and characteristic wavelengths. These models are gray wolf optimized support vector regression (GWO-SVR), particle swarm optimized support vector regression (PSO-SVR), and support vector regression-based integrated learning (AdaBoost-SVR). Comparing the models showed that the AdaBoost-SVR model, utilizing the Eigen wavelengths extracted through the CARS-SPA and CARS algorithms, stood out as the most accurate in predicting wheat protein and fat contents, with Rp 2 values of 0.9789 and 0.9651, respectively. The study findings indicate that combining HSI with AdaBoost-SVR integrated learning modeling enables rapid, non-destructive determination of wheat protein and fat contents with greater accuracy than using a single machine learning (ML) model.