No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Evaluating the Dry Matter Content of Raw Yams Using Hyperspectral Imaging Spectroscopy and Machine Learning
Abstract
Yams (Dioscorea spp.) are important food and commercial crops in West African countries. They contribute significantly to global food production and provide dietary energy. The quality of yam food products depends on specific internal and external parameters, such as the DMC and other biochemical traits. However, measuring these traits can be challenging, particularly when analyzing many genotypes. This study aimed to evaluate the feasibility of using near-infrared (NIR) hyperspectral imaging (932–1721 nm) along with machine learning to rapidly measure the dry matter content (DMC) of fresh, intact yam tubers. Hyperspectral images were acquired across the yam tuber’s cross-sections, and the resulting spectra from the images were averaged and preprocessed. Partial least square regression (PLSR) combined with successive progressions algorithms (SPA), Competitive Adaptive Reweighted Sampling (CARS), Artificial Neural network (ANN) and Boruta algorithms (BA) were used
to select the important wavelengths for developing a prediction model for DMC (g/100 g). The PLSR-SPA-CARS model showed the most accurate prediction performances with a coefficient of determinations in calibration (R2cal) and prediction (R2pred) of 0.974 and 0.958, respectively, and low root mean square error (RMSEP) of 0.898 g/100 g. The distribution of DMC was visually represented by projecting the developed model to generate
color chemical maps. This study resolves that NIR hyperspectral imaging can rapidly assess the DMC of fresh, intact yam tubers.
... However, measuring these parameters is often time-consuming and laborious. Given the relationship between light absorption and leaf physiological substances such as pigments [11] , water [12] , and dry matter [13] , remote sensing methods hold great potential for quickly accessing photosynthetic differences among plant species [14] or genotypes [15] . ...
Effectively evaluating and estimating the photosynthetic capacities of different poplar genotypes is essential for selecting and breeding poplars with high productivity. This study measured leaf hyperspectral reflectance, net photosynthetic rate (Pn), transpiration rate (Tr), intercellular CO2 concentration (Ci), and stomatal conductance (Gs) across the upper-, middle- and lower-layer leaves of six poplar genotypes. Photosynthetic capacities and spectral differences were assessed among these genotypes. By analyzing the correlation of photosynthetic parameters and spectral characteristics, the photosynthetic parameters were also estimated from hyperspectral parameters using BP neural networks. Significant differences were observed in the photosynthetic parameters among six poplar genotypes. Populus tremula × P. alba exhibited the highest photosynthetic rate, while Populus hopeiensis showed the lowest. Leaves in the middle layer demonstrated greater photosynthetic capacities than those in the other layers. Leaf reflectance among the six poplar genotypes differed significantly in the ranges of 400−760 nm, 800−1,300 nm, 1,500−1,800 nm, and 1,900−2,000 nm. Values for MTCI, WI, REP, PRI, and first-order derivative at 891 nm also showed significant differences. Hyperspectral parameters, including first-order derivative spectra (FDS), raw spectral reflectance, and photosynthetic parameters, showed strong correlations in the red light (670 nm), near-infrared (760−940 nm), and short-wave infrared (1,800−2,500 nm). Four photosynthetic parameters including Pn, Tr, Ci, and Gs were estimated using BP neural network models and R² were 0.56, 0.44, 0.35, and 0.35, respectively. The present results indicate that hyperspectral reflectance can effectively distinguish between different poplar genotypes and estimate photosynthetic parameters, highlighting its great potential for studying plant phenomics.
The study investigated the use of the near-infrared hyperspectral imaging (NIR-HSI) technique (932 – 1721 nm) to rapidly evaluate the starch, sugar, and amylose content of fresh, intact yam tubers and the textural qualities of boiled yam. These quality characteristics often influence consumers’ and farmers’ acceptance of new yam varieties. Traditional methods for their determination are expensive, time-consuming, and sometimes subjective. The NIR-HSI system combined with three Effective Wavelengths (EWs) selection algorithms, including Successive
Projections Algorithms (SPA), Competitive Adaptive Reweighted Sampling (CARS), and Boruta Algorithm (BA), was used to extract the important spectral features. The PLSR-SPA-CARS gave the best prediction models in most cases, with a coefficient of determination in prediction (R2 pre) of 0.952 for starch, 0.935 for sugar, and 0.978 for amylose content, respectively. The spatial distribution of starch, sugar, and amylose was visualized using the
optimized PLSR model. Additionally, PLSR-SNV-SG (Standard Normal Variate and Savitzsky-Golay) showed the best R2 pred of 0.846 for peak force (hardness) and 0.538 for the area under the curve (chewiness) of boiled yam. This study has demonstrated the potential of NIR-HSI techniques to rapidly predict the quality of fresh yam and its boiled food product.
The SOP describes standard method for evaluation of cooking time, optimum cooking time and instrumental texture of boiled yam by penetration and TPA techniques.
Accurate assessment of isoflavone and starch content in Puerariae Thomsonii Radix (PTR) is crucial for ensuring its quality. However, conventional measurement methods often suffer from time-consuming and labor-intensive procedures. In this study, we propose an innovative and efficient approach that harnesses hyperspectral imaging (HSI) technology and deep learning (DL) to predict the content of isoflavones (puerarin, puerarin apioside, daidzin, daidzein) and starch in PTR. Specifically, we develop a one-dimensional convolutional neural network (1DCNN) model and compare its predictive performance with traditional methods, including partial least squares regression (PLSR), support vector regression (SVR), and CatBoost. To optimize the prediction process, we employ various spectral preprocessing techniques and wavelength selection algorithms. Experimental results unequivocally demonstrate the superior performance of the DL model, achieving exceptional performance with mean coefficient of determination (R²) values surpassing 0.9 for all components. This research underscores the potential of integrating HSI technology with DL methods, thereby establishing the feasibility of HSI as an efficient and non-destructive tool for predicting the content of isoflavones and starch in PTR. Moreover, this methodology holds great promise for enhancing efficiency in quality control within the food industry.
Background:
Chayote is an underutilized species of Cucurbitaceae. It is rich in nutrients such as protein, minerals, phenols and its extracts have anti-cardiovascular and anti-cancer effects, making it a versatile plant for both medicinal and culinary purposes. Although research on its root tuber is limited, they are rich in starch and have a structure similar to that of potatoes, cassava, and sweet potatoes. Therefore, they can serve as potential substitutes for potatoes and offer promising prospects as agricultural and industrial resources. However, the physiological and cellular mechanisms of chayote root tuber formation and development are still unclear.
Results:
In this study, we observed the growth habit of 'Tuershao' (high yield of root tuber). The results revealed that the tuber enlargement period of 'Tuershao' lasts approximately 120 days, with the early enlargement phase occurring during 0-30 days, rapid enlargement phase during 30-90 days, and maturation phase during 90-120 days. Physiological indicators demonstrated a gradual increase in starch content as the tuber developed. The activities of sucrose synthase (SUS) and invertase (VIN) showed a consistent trend, reaching the highest level in the rapid expansion period, which was the key enzyme affecting tuber expansion. Moreover, the special petal like structure formed by the secondary phloem and secondary xylem of the tuber resulted in its enlargement, facilitating the accumulation of abundant starch within the thin-walled cells of this structure. Principal component analysis further confirmed that starch content, SUS and VIN activities, as well as the concentrations of calcium (Ca), iron (Fe), and selenium (Se), were the major factors influencing tuber development. Moreover, the low temperature environment not only promoted the growth of 'Tuershao' tubers but also enhanced the accumulation of nutritional substances.
Conclusions:
These findings contribute to a deeper understanding of the formation and developmental mechanisms of 'Tuershao' tubers, providing valuable guidance for cultivation practices aimed at improving crop yield.
BACKGROUND
Consumer preferences for boiled or fried pieces of roots, tubers and bananas (RTBs) are mainly related to their texture. Different raw and cooked RTBs were physiochemically characterized to determine the effect of biochemical components on their cooking properties.
RESULTS
Firmness in boiled sweetpotato increases with sugar and amylose contents but no significant correlation was observed between other physicochemical characteristics and cooking behaviour. Hardness of boiled yam can be predicted by dry matter (DM) and galacturonic acid (GalA) levels. For cassava, no significant correlation was found between textural properties of boiled roots and DM, but amylose and Ca²⁺ content were correlated with firmness, negatively and positively, respectively. Water absorption of cassava root pieces boiled in calcium chloride solutions was much lower, providing indirect evidence that pectins are involved in determining cooking quality. A highly positive correlation between textural attributes and DM was observed for fried plantain, but no significant correlation was found with GalA, although frying slightly reduced GalA.
CONCLUSION
The effect of main components on texture after cooking differs for the various RTBs. The effect of global DM and major components (i.e. starch, amylose) is prominent for yam, plantain and sweetpotato. Pectins also play an important role on the texture of boiled yam and play a prominent role for cassava through interaction with Ca²⁺. © 2023 Bill and Melinda Gates Foundation. Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
The flavor of Pomelo is highly variable and difficult to determine without peeling the fruit. The quality of pomelo flavor is due largely to the total soluble solid content (TSSC) in the fruit and there is a commercial need for a quick but nondestructive TSSC detection method for the industrial grading of pomelo. Due to the large size and thick mesocarp of pomelo, determining the internal quality of a pomelo fruit in a nondestructive manner is difficult, and the detection accuracy is further complicated by the noise typically generated by the common methods for the internal quality detection of other fruits. Thus, the aim of this study was to determine the optimal method to accurately detect pomelo TSSC and find a de-noising model which reduces the influence of noise on the optimal method’s results. After developing a full-transmission visible/near infrared (VIS/NIR) spectroscopy sampling method, the confirming experimental results showed that the optimal pomelo TSSC detection model was Savitzky Golay + standard normal variate + competitive adaptive reweighted sampling + partial least squares regression. The R² and RMSE of the calibration set for pomelo TSSC detection were 0.8097 and 0.8508, respectively, and the R² and RMSE of the validation set for pomelo TSSC detection were 0.8053 and 0.8888, respectively. Both reference and dark de-noising are important for pomelo internal quality detection and should be calibrated frequently to compensate for time drift. This study found that large sensor response translation noise can be reduced with an artificial horizontal shift. Data supplementation is efficient for improving the adaption of the detection model for batch differences in pomelo samples. Using this optimized de-noising model to compensate for time drift, sensor response translation, and batch differences, the developed detection method is capable of satisfying the requirements of the industry (TSSC detection R² was equal or larger than 0.9, RMSE was less than 1). These results indicate that full-transmission VIS/NIR spectroscopy can be exploited to realize the nondestructive detection of pomelo TSSC on an industrial scale, and that the methodologies used in this study can be immediately implemented in real-world production.
Potato–wheat flour mixture (PWFM) has important significance for the development of potato staple food products. However, conventional chemical methods are unable to detect the content of potato flour. Hyperspectral imaging (HSI) technology combined with chemometrics was investigated to predict the feasibility of potato flour content in PWFM. In this study, seven pretreatment algorithms were performed to process the raw spectral data. The results showed that the standard normalised variables (SNV) to establish a partial least square regression (PLSR) model had the most predictive accuracy with the determination coefficient of prediction (Rp2) of 0.9729. The characteristic wavelengths of raw spectral information were extracted using three algorithms to reduce the model complexity. Eventually, the extracted characteristic wavelengths of potato flour content combine with SNV to establish PLSR, principal component regression (PCR) and support vector regression (SVR) models. The results revealed that the optimal model was SNV‐competitive adaptive reweighted sampling (CARS)‐PLSR with the values of Rp2 of 0.9853. These results showed that HSI determination of potato flour content in PWFM was feasible and provided a non‐destructive method.
Hyperspectral imaging (HSI) is one of the most often used techniques for rapid quality evaluation for various applications. It is a non-destructive technique that effectively evaluates the quality attributes of root and tuber crops, including yam and cassava, and their food products. Hyperspectral imaging technology, which combines spectroscopy and imaging principles, has an advantage over conventional spectroscopy due to its ability to simultaneously evaluate the physical characteristics and chemical components of various food products and specify their spatial distributions. HSI has demonstrated significant potential for obtaining quick information regarding the chemical composition of the root and tuber, such as starch, protein, dry matter, amylose, and soluble sugars, as well as physical characteristics such as textural properties and water binding capacity. This review highlights the principles of near-infrared hyperspectral imaging (NIR-HSI) techniques combined with relevant image processing tools. It then provides cases of its application in determining crucial biochemical quality traits and textural attributes of roots and tuber crops, focusing on cassava and yam. The need for more information on using NIR-HSI in the quality evaluation of yam and cassava was underscored. It also presents the challenges and prospects of this technology.
BACKGROUND
The purpose of this study was to investigate the potential of hyperspectral imaging for the characterization of cooking quality parameters, dry matter content (DMC), water absorption (WAB), and texture in cassava genotypes contrasting for their cooking quality.
RESULTS
Hyperspectral images were acquired on cooked and fresh intact longitudinal and transversal slices from 31 cassava genotypes harvested in March 2022 in Colombia. Different chemometric methods were tested for the quantification of DMC, WAB, and texture parameters. Data analysis was conducted through partial least squares regression, K nearest neighbors regression, support vector machine regression and CovSel multiple linear regression (CovSel_MLR). Efficient performances were obtained for DMC using CovSel_MLR with, coefficient of multiple determination Rp2=0.94, root‐mean‐square error of prediction RMSEP = 0.96 g/100 g, and ratio of the standard deviation values RPD = 3.60. High heterogeneity was observed between contrasting genotypes. The predicted distribution of DMC within the root can be homogeneous or heterogeneous depending on the genotype. Weak predictions were obtained for WAB and texture parameters.
CONCLUSIONS
This study showed that hyperspectral imaging could be used as a high‐throughput phenotyping tool for the visualization of DMC in contrasting cooking quality genotypes. Further improvement of protocols and larger datasets are required for WAB and texture quality traits. © 2023 The Authors. Journal of The Science of Food and Agriculture published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.
Potatoes are generally consumed directly as a staple food or used for processing, depending on the quality of raw materials. Dry matter content (DMC) is the most critical characteristic of potatoes, as it determines the processing and the final product quality. This study aimed to investigate the potential of different optical sensing systems in predicting the DMC of intact potato tubers, and the efficacy of classifying potatoes based on dry matter levels. The whole tubers were scanned using three optical sensing modes (transmittance spectra, interactance spectra and hyperspectral imaging). PLSR and different classifiers (PLSDA, SVM and ANN) were utilized to build the prediction and classification models, respectively. To extract the most influential wavelengths related to the prediction of DMC, the CARS and CSMW techniques were applied. The results indicated that the DMC of two sections on the equator of the potato tuber belly can well represent the DMC of the intact potato, and together with the spectral detection at the equatorial position, it provided good performance. The CARS-PLSR prediction model in transmittance mode showed stronger correlations than other systems, with Rp and RMSEP values of 0.968 and 0.413%, respectively. The CARS-SVM-Linear classification model exhibited the best performance with classification rates of 100% and 97.62% in the training and testing sets, respectively. Moreover, the spectra preferred by CARS and CSMW variable selection methods in transmittance mode overlapped near the absorption peak at 980 nm, indicating the importance of this band for predicting DMC. This study presented the feasible application of using spectroscopy to evaluate the DMC of intact potatoes and classify potatoes based on thresholds that are crucial to consumers and food processors.
Graphical abstract
Abstract This study investigated the hyperspectral reflectance response of time series generated during oven drying to changes in the moisture content of potato tubers. Seventeen preprocessing methods were used to eliminate the influence of spectral noise on the spectral characteristic curve. Algorithms such as CatBoost, LightGBM, and XGBoost are used to obtain the first 40 effective characteristic spectra of hyperspectral images, which reduces the redundancy of data and improves the prediction accuracy. The water content prediction model of potato tubers was established by using the selected characteristic bands. The results showed that the combined model based on Lasso and XGBoost algorithm had the strongest prediction ability. The best model is MF-Lasso-XGBoost, which has R2 value of 0.8908, Rmse of 0.0610, Mdae of 0.0389, and R2cv of 0.8448. This research can provide reference for the detection of potato moisture content and theoretical basis for the development of crop moisture detector.
The primary objective of this study was to assess the effects of genotype (G), location (L), and G × L interaction on tuber quality traits (dry matter content and starch pasting parameters) in white Guinea yam (Dioscorea rotundata Poir.). Variability in tuber dry matter and starch pasting properties was examined using 18 advanced breeding lines and two dominant landrace cultivars of white Guinea yam grown in three different agroecological zones (forest‒savanna transition, southern Guinea savanna, and rainforest) in Nigeria. The starch pasting properties were evaluated using a Rapid Visco Analyser. Our results show that the G × L interaction effect was low compared to the genotype and growing location effects on the variability of key starch properties. In addition, the repeatability of trait performance across locations was high and relatively uniform for key traits, suggesting that any of the three locations used in this study can be employed for their evaluation. Furthermore, TDr1100873 had a higher dry matter content than the dominant landrace cultivars (Amula and Meccakusa) but was similar to them in starch pasting properties. Hence, TDr1100873 is considered a suitable variety for future breeding activities.
Starch is an important quality index in potato, which contributes greatly to the taste and nutritional quality of potato. At present, the determination of starch depends on chemical analysis, which is time consuming and laborious. Thus, rapid and accurate detection of the starch content of potatoes is important. This study combined hyperspectral imaging with chemometrics to predict potato starch content. Two varieties of Kexin No.1 and Holland No.15 potatoes were used as experimental samples. Hyperspectral data were collected from three sampling sites (the top, umbilicus, and middle regions). Standard normal variate (SNV) was used for spectral preprocessing, and three different methods of competitive adaptive reweighted sampling (CARS), iterative variable subset optimization (IVSO), and the variable iterative space shrinkage approach (VISSA) were used for characteristic wavelength selection. Linear partial least‐squares regression (PLSR) and nonlinear support vector regression (SVR) models were then established. The results indicated that the sampling site has a considerable impact on the accuracy of the prediction model, and the umbilicus region with CARS‐SVR model gave best performance with correlation coefficients in calibration (Rc) of 0.9415, in prediction (Rp) of 0.9346, root mean square errors in calibration (RMSEC) of 15.9 g/kg, in prediction (RMSEP) of 17.4 g/kg, and residual predictive deviation (RPD) of 2.69. The starch content in potatoes was visualized using the best model in combination with pseudo‐color technology. Our research provides a method for the rapid and nondestructive determination of starch content in potatoes, providing a good foundation for potato quality monitoring and grading. In this paper, the starch content in different positions of potato was predicted and visualized.
In the era of healthcare, and its related research fields, the dimensionality problem of high dimensional data is a massive challenge as it contains a huge number of variables forming complex data matrices. The demand for dimension reduction of complex data is growing immensely to improvise data prediction, analysis and visualization. In general, dimension reduction techniques are defined as a compression of dataset from higher dimensional matrix to lower dimensional matrix. Several computational techniques have been implemented for data dimension reduction, which is further segregated into two categories such as feature extraction and feature selection. In this review, a detailed investigation of various feature extraction and feature selection methods has been carried out with a systematic comparison of several dimension reduction techniques for the analysis of high dimensional data and to overcome the problem of data loss. Then, some case studies are also cited to verify the better approach for data dimension reduction by considering few advances described in the technical literature. This review paper may guide researchers to choose the most effective method for satisfactory analysis of high dimensional data.
Background
The accurate estimation of potato yield at regional scales is crucial for food security, precision agriculture, and agricultural sustainable development.
Methods
In this study, we developed a new method using multi-period relative vegetation indices (rVIs) and relative leaf area index (rLAI) data to improve the accuracy of potato yield estimation based on the weighted growth stage. Two experiments of field and greenhouse (water and nitrogen fertilizer experiments) in 2018 were performed to obtain the spectra and LAI data of the whole growth stage of potato. Then the weighted growth stage was determined by three weighting methods (improved analytic hierarchy process method, IAHP; entropy weight method, EW; and optimal combination weighting method, OCW) and the Slogistic model. A comparison of the estimation performance of rVI-based and rLAI-based models with a single and weighted stage was completed.
Results
The results showed that among the six test rVIs, the relative red edge chlorophyll index (rCIred edge) was the optimal index of the single-stage estimation models with the correlation with potato yield. The most suitable single stage for potato yield estimation was the tuber expansion stage. For weighted growth stage models, the OCW-LAI model was determined as the best one to accurately predict the potato yield with an adjusted R² value of 0.8333, and the estimation error about 8%.
Conclusion
This study emphasizes the importance of inconsistent contributions of multi-period or different types of data to the results when they are used together, and the weights need to be considered.
Visible and near infrared (Vis-NIR) hyperspectral imaging was used for fast detection and visualization of soluble solid content (SSC) in ‘Beijing 553’ and ‘Red Banana’ sweet potatoes. Hyperspectral images were acquired from 420 ROIs of each cultivar of sliced sweet potatoes. There were 8 and 10 outliers removed from ‘Beijing 553’ and ‘Red Banana’ sweet potatoes by Monte Carlo partial least squares (MCPLS). The optimal spectral pretreatments were determined to enhance the performance of the prediction model. Successive projections algorithm (SPA) and competitive adaptive reweighted sampling (CARS) were employed to select characteristic wavelengths. SSC prediction models were developed using partial least squares regression (PLSR), support vector regression (SVR) and multivariate linear regression (MLR). The more effective prediction performances emerged from the SPA–SVR model with Rp² of 0.8581, RMSEP of 0.2951 and RPDp of 2.56 for ‘Beijing 553’ sweet potato, and the CARS–MLR model with Rp² of 0.8153, RMSEP of 0.2744 and RPDp of 2.09 for ‘Red Banana’ sweet potato. Spatial distribution maps of SSC were obtained in a pixel-wise manner using SPA–SVR and CARS–MLR models for quantifying the SSC level in a simple way. The overall results illustrated that Vis-NIR hyperspectral imaging was a powerful tool for spatial prediction of SSC in sweet potatoes.
The review aimed to identify the different high‐throughput phenotyping (HTP) techniques that used for quality evaluation in cassava and yam breeding programmes, and this has provided insights towards the development of metrics and their application in cassava and yam improvements. A systematic review of the published research articles involved the use of NIRS in analysing the quality traits of cassava and yam was carried out, and Scopus, Science Direct, Web of Sciences and Google Scholar were searched. The results of the review established that NIRS could be used in understanding the chemical constituents (carbohydrate, protein, vitamins, minerals, carotenoids, moisture, starch, etc.) for high‐throughput phenotyping. This study provides preliminary evidence of the application of NIRS as an efficient and affordable procedure for HTP. However, the feasibility of using mid‐infrared spectroscopy (MIRS) and hyperspectral imaging (HSI) in combination with the NIRS could be further studied for quality traits phenotyping.
Mid-infrared (MIR) microscopy provides rich chemical and structural information about biological samples, without staining. Conventionally, the long MIR wavelength severely limits the lateral resolution owing to optical diffraction; moreover, the strong MIR absorption of water ubiquitous in fresh biological samples results in high background and low contrast. To overcome these limitations, we propose a method that employs photoacoustic detection highly localized with a pulsed ultraviolet laser on the basis of the Grüneisen relaxation effect. For cultured cells, our method achieves water-background suppressed MIR imaging of lipids and proteins at ultraviolet resolution, at least an order of magnitude finer than the MIR diffraction limits. Label-free histology using this method is also demonstrated in thick brain slices. Our approach provides convenient high-resolution and high-contrast MIR imaging, which can benefit the diagnosis of fresh biological samples.
This study was conducted to determine the changes in dry matter (DM), free sugar (FS) and starch contents in the three yam (D. rotundata) tuber regions under two storage environmental conditions (natural light-yam barn, or and darkness). Three varieties of D. rotundata (alaako, Dodoro and Odo) were studied. The dry matter content of the Head region was the highest (28.6%), followed by the Middle (26.9%) and Tail (22.3%) regions. Change in dry matter content of the tuber is mainly due to rapid changes in the dry matter content of the Tail region over time. Storage environment significantly (P< 0.05) affected dry matter content of the three tuber regions at 8 and 16 weeks in storage (WIS). Head regions per variety were lower in dry matter when stored under Light than in the Dark at 8 WIS, while the reverse was the case for the Tail. This may relate to the release of endodormancy. By 16 WIS, when tubers stored under Light had well developed sprouts and those stored in the Dark only showed the first sign of sprouting, all Head regions had higher dry matter content under Light than in Dark storage. This may be due to the presence of sprouts. Storage environment did not significantly affect starch content at the various tuber regions. Head regions had higher percentage free sugar content under Light than Dark storage, and these effects were manifested between 8 and 16 WIS.
Visible/near-infrared (VIS/NIR) hyperspectral reflectance imaging was evaluated as a technique toward rapid prediction of the glucose and sucrose percentages in two common fresh use and chipping potato cultivars. Tubers were sampled and held in multiple storage temperatures in an attempt to develop uniform and broad constituent distributions. Each tested sample was a 12.7-mm-thick slice cut uniformly from all tubers. Multiple features were extracted from samples including mean reflectance spectra and curve feature parameters yielded from an exponential model. Both glucose and sucrose ratios were measured using the Megazyme sucrose/D-glucose assay procedure as reference values. Partial least squares regression (PLSR), feed forward neural networks (FFNN), radial basis functions neural networks (RBFNN), and exact design radial basis functions (RBFNNE) neural networks were used for building calibration and prediction models. PLSR results demonstrated strongly correlated models built using mean reflectance spectra for glucose for Russet Norkotah (RN) with R (RPD) (or correlation coefficient (ratio of sample standard deviation to standard error of prediction)) values as high as 0.97 (3.58), whereas those values were 0.81 (1.70) for Frito Lay 1879 (FL). Sucrose models showed less correlation performance with R (RPD) values as high as 0.60 (1.14) for FL and 0.38 (1.00) for RN. Wavelength selection/prediction using interval partial least squares (IPLS) and genetic algorithm (GA) was conducted on the data, and PLSR and NN results were close to the full-wavelength models for the glucose and sucrose of both cultivars with a preference given to IPLS as it yields less selected variables than GA. Applying K-nearest neighbor (Knn) and partial least squares discriminant analysis (PLSDA) on mean reflectance spectra resulted in glucose misclassification errors of 14 % and 18 % for FL and RN, respectively. However, classification errors were higher for sucrose indicating lower accuracy for this sugar (34 and 30 % for FL and RN). The results in this study give promise to the possibility of measuring each of these sugars rapidly for quality control and monitoring in the potato industry.
Visible/Near-infrared (Vis/NIR) spectroscopy is widely used in the detection of dry matter content (DMC) of potatoes. However, biological variability (e.g., cultivar and season) will affect the potato DMC and spectral features, and will further cause the DMC prediction model ineffective. This study aimed to develop robust Vis/NIR models for predicting potato DMC with influence of cultivar and season. The local and global models were developed to explore the influence of cultivar and season. The Mahalanobis distance and concentration gradient (MD-CG) method was developed to select representative samples, and the combinations of different variable selection methods (CARS, SPA and CSMW) and model updating methods (SBC and recalibration) were investigated for model enhancement. The results indicated that 10 new samples selected by MD-CG method, combined with variable selection and model updating, were sufficient to improve the performance of the local (RPDp>1.7) and global (RPDp>2) models. In the local models, for the datasets with different cultivars (EG-2021, XS-2021 and AT-2021), the optimal results were obtained using CSMW combined with recalibration, and the RMSEp was decreased from 4.18%, 1.14%, 2.54–1.05%, 0.72%, 0.79%, respectively. For the datasets with different seasons (FA-2022), the optimal result was obtained by using SPA combined with recalibration, and the RMSEp was decreased from 3.70% to 0.91%. For the global model, CSMW combined with recalibration and SPA combined with SBC obtained better results, with RMSEp decreasing from 0.83% to 0.52% and 0.51%, respectively. The MD-CG method and the combinations of variable selection and model updating proposed in this study are important to reduce the influence of external conditions and enhance the model robustness to biological variability.
Mutton kebab is an attractive type of meat product with high nutritional value, and is favored by consumers worldwide. However, mutton kebab is often subjected to adulteration due to its high price. Chicken, duck, and pork are frequently used as adulterated substitutes. The purpose of current study aims at developing a methodology based on hyperspectral imaging (HSI, 400–1000 nm) for identifying the authenticity of fresh and cooked mutton kebabs. Kebab samples were individually scanned using HSI system in their fresh and cooked states. Spectra of chicken, duck, pork, and mutton kebabs were first extracted from representative regions of interest (ROIs) identified in their calibrated hyperspectral images. After that, principal component analysis (PCA) was carried out, and results showed that the first three or two PCs were effective for identifying fresh or cooked samples of different meat species. Different effective modeling algorithms including k-nearest neighbor (KNN), partial least squares discriminant analysis (PLS-DA), and support vector machine (SVM) algorithms combined with different preprocessing methods were employed to develop classification models. Performances exhibited that PLS-DA models using raw spectra outperformed the KNN and SVM models, and the accuracies reached both 100 % in prediction sets for fresh and cooked meat kebabs, respectively. Moreover, compared to iteratively variable subset optimization (IVSO), random frog (RF), and successive projections algorithm (SPA) algorithms, the PC loadings successfully screened 14 and 8 effective wavelengths for fresh and cooked meat kebabs, respectively, from the complex original full-band wavelengths. The PC-PLS-DA models showed the optimal predicted performances with overall classification accuracies of 97.5 % and 100 %, sensitivity values of 1.00 and 1.00, specificity values of 0.97 and 1.00, precisions of 0.91 and 1.00, for fresh and cooked mutton kebabs, respectively. Furthermore, the visualization of classification maps confirmed the experimental results intuitively. Overall, it was evident that HSI showed immense potential to identify the authenticity of fresh and cooked mutton kebabs when substituted by different meats including chicken, duck, and pork.
Near-infrared spectroscopy (NIRS) and hyperspectral imaging (HSI) techniques combined with chemometric method are emerging techniques and have been studied and applied to address current challenges in anaerobic digestion (AD) plants, such as the heterogeneity of feedstocks, low methane yield, process instability and digestate management. Prior to the AD process, the rapid and accurate measurement of the feedstocks’ chemical composition can predict the biochemical methane potential and discern the potential microorganism inhibitors. During the AD process, monitoring the intermediate products in the AD digesters by using NIRS or HSI techniques allows for process optimization and avoids potential AD failure. Regarding digestate management, the NIRS or HSI can be applied to determine the biological stability and evaluate the digestate quality. In this review, we summarize recent research advances in monitoring AD process parameters and quality of feeding substrate and digestate using NIRS and HSI combined with machine learning techniques. This review highlights the application of NIRS and HSI technology in the AD of organic wastes with particular emphasis on the application drawbacks and possible enhancement solutions. In general, the existing machine learning augmented NIRS can obtain satisfactory quantification results. Future researches on characterization of high-moisture heterogeneous substrate and real-time monitoring AD by HSI combined deep learning are still in demand.
Spectroscopy is essential to understand a series of phenomena in multiple fields of study. In remote sensing, vegetation analysis is one of the most prominent fields to explore, aiming to improve a specific task. As a task, modeling insect damage in the plants is essential to establish the correct management of agricultural farmlands. Hyperspectral data, which can be acquired with field spectroscopy at plant or leaf level, is a non-direct, rapid, and trustworthy approach to indicate its health. However, the spectral redundancy inherent is a challenge for the information extraction process, making the pre-processing phase an essential part of the analysis. Currently, artificial intelligence techniques, mostly based on machine and deep learning methods, are a standard application in data processing, being pre-processing techniques an essential part of it. But few studies aimed to measure the impact of such processes in vegetation monitoring, specifically with insect damage and spectral data. Here, we provide an analysis of the impact of pre-processing techniques on machine learning algorithms’ performance over said classification task. For this, we used a field spectroradiometer that operates within the 350-1,000 nm and 1,000-2,500 nm ranges. The dataset was composed of multiple spectral measurements that took place on different days in a controlled environment with soybean plants. As pre-processing techniques, methods like baseline removal, smoothing, first and second-order derivatives, standard normal variate (SNV), multiplicative scatter correction (MSC), and principal components analysis (PCA) were investigated. Several machine learning algorithms and one deep learning method were applied to model the datasets. The impact of the pre-processing techniques was measured within validation metrics relate to its accuracy. Our results indicated that the Extra-Tree (ExT) algorithm was better, mainly when first-order derivative data were extracted from the dataset (accuracy equal to 93.68%). A ranking approach indicated that the most contributive spectral region situates at the near-infrared, between 784 and 911 nm. Our investigation also demonstrates that a deep neural network (DNN) did not return a satisfactory result over raw reflectance data. However, when considering a combination of PCA over the 2nd derivative data, it achieved similar results to the ExT algorithm (accuracy of 91.95%). The implications of such, alongside the ranking approach, are discussed in this paper. We hope that the information presented here serves as a framework for future research when applying pre-processing techniques alongside the machine and deep learning methods over spectral data.
This study used a portable near-infrared (NIR) spectrometer at wavelengths of 570–1031 nm to evaluate starch content (SC) and dry matter content (DMC) in fresh cassava tubers. An improved model was developed for the prediction of cassava tuber quality. The cassava samples were taken from four main varieties: CMR38-125-77, KU50, RY11, and RY9. The samples were obtained 4–12 months after planting (MAP). Partial least squares (PLR) regression was combined with different variable selection methods and spectral pre-treatment. Their accuracies were then compared. Variable selection methods included the successive projections algorithm (SPA) and the genetic algorithm (GA). The NIR spectra were obtained in the interactance mode under field conditions. The GA wavelengths combined with sequential pre-processing by orthogonalization (SPORT) pre-processing provided the optimum model for predicting both the SC and the DMC of cassava. The R²p, RMSEp, and RPD of SC were 0.91, 1.76%, and 3.26, respectively, and those of DMC were 0.75, 2.00%, and 2.00, respectively. The most effective model was tested against unknown samples of newly developed varieties obtained from different harvest seasons, yielding RMSEp and bias values of 2.37% and −9.178 × 10⁻⁶%, respectively, for SC. For DMC, the RMSEp and bias values were 2.67% and 4.16 × 10⁻¹⁴%, respectively. The results suggest that the calibration model could be used to monitor the internal quality of cassava tubers in the field. The variety, age, position, and section of the tubers had a slight influence on the prediction performance; however, the prediction accuracy was acceptable for in-field applications. The in-field portable NIR spectrometer could become a new tool for breeders, saving time and costs. Breeders could evaluate SC without destroying the cassava roots or stalks and could correct and inspect the behaviour of the SC and DMC accumulation.
In order to realize rapid and non-destructive detection of hardness for maize kernels, a method for quantitative hardness measurement was proposed based on hyperspectral imaging technology. Firstly, the regression model of hardness and moisture content was established. Then, based on reflectance hyperspectral imaging at wavelengths within 399.75-1005.80 nm, the prediction model of the moisture content was studied by the partial least squares regression (PLSR) based on the characteristic wavelengths, which was selected through successive projection algorithm (SPA). Finally, the hardness prediction model was validated by combing the prediction model of moisture content with the regression model of hardness. The coefficient of determination (R²), the root mean square error (RMSE) the ratio of performance-to-deviation (RPD) and the ratio of error range (RER) of hardness prediction were 0.912, 17.76 MPa, 3.41 and 14, respectively. Therefore, this study provided a method for rapid and non-destructive detection of hardness of maize kernels.
Partial least squares regression (PLSR) modeling was performed to predict the moisture content in steamed, dried purple sweet potato based on spectral data obtained from hyperspectral imaging analysis. The PLSR model with a combination of multiplicative scatter correction, Savitzky–Golay, and first derivative exhibited the highest accuracy (RP2 = 0.9754). The wavelengths found that strongly affected the PLSR model were 961.12, 1065.50, 1083.93, 1173.23, and 1233.89 nm. These wavelengths were associated with the O–H second overtone and the second overtone of C–H, C–H2, and C–H3. When PLSR modeling was performed using these selected wavelengths, the prediction accuracy of the PLSR model exhibited high accuracy (RP2 = 0.9521). Therefore, the moisture content could be predicted with high accuracy using only five wavelengths rather than the full spectrum.
Total anthocyanin (TA) and moisture content (MC) are critical indexes of processed purple sweet potatoes (PSPs). The current study examined the feasibility of hyperspectral imaging to investigate how MC and TA contents of PSPs change during processing under two different drying methods namely convective hot‐air drying (CHD) and microwave drying (MD). Models based on spectral data and the integration of spectral and one or three gray‐level co‐occurrence matrix features were established with partial least square regression. The best prediction outcome based on Spectra + HOMOGENEITY + CONTRAST + ENTROPY combinations were ( = 0.862, 0.847; RMSEP = 0.079, 0.303), respectively, for MC and TA for the CHD samples. Similarly, for the MD samples, the best prediction outcomes were ( = 0.867, 0.859; RMSEP = 0.088, 0.241), respectively, for MC and TA. Image algorithms were also developed to generate distribution of MC and TA in some representative samples. Anthocyanins have been reported as having the potential to lower blood pressure, improve eyesight, reduce cancer cell multiplication, impede tumor formation, and prevent diabetes. Studies have, however, shown that cooking modes including boiling, baking, and steaming decreased the contents of anthocyanins. This study assessed the impact of convective hot‐air and microwave drying (MD) process of purple sweet potato (PSP) on total anthocyanin (TA) content as a function of moisture loss. Hyperspectral Imaging (HSI) was useful in estimating not only the contents of TA and moisture, but also in generating a visual map of their distribution pattern in the processed PSPs. The TA content of PSPs processed by MD was higher than that by convective hot‐air drying, but the distribution uniformity of TA in microwave dried PSPs was worse than that in convective hot‐air dried samples. Finally, to guarantee consistency in drying processes of food products, the HSI could be used to obtain visual images of their chemical parameters.
The feasibility of using NIR hyperspectral imaging technique for predicting fat and moisture contents in salmon fillets was assessed by integrating both characteristic wavelengths and image texture features. Monte Carlo-uninformative variable elimination (MC-UVE) and successive projections algorithm (SPA) were combined to extract characteristic wavelengths. Ten textural features of the principal component images were obtained using histogram statistics (HS) and gray level co-occurrence matrices (GLCMs) methods. Three types of models (PLS, MLR and LS-SVM) were established based on different types of inputs including only characteristic wavelengths, only texture parameters and combination both characteristic wavelengths and textures, respectively. Compared among all models, LS-SVM model coupled with wavelength and texture information gave the highest prediction accuracies with RP = 0.9685, RMSEP = 1.1750, RPD = 4.0162 for fat and RP = 0.9688, RMSEP = 0.8021, RPD = 4.0357 for moisture, respectively. This study showed that the prediction accuracy can be improved by combining spectral features with textural features and the fusion of characteristic wavelength and textural features had better potential than single spectral information in assessing the fat and moisture contents of salmon. Satisfactory prediction results confirmed the suitability of NIR hyperspectral imaging for quantitative prediction of fat and moisture in salmon.
Freezing, heating, and pickling are common processes for pork meats. Unsaturated fatty acids including monounsaturated fatty acids and polyunsaturated fatty acids are indispensable nutrition beneficial to human’s health and growth. However, Unsaturated fatty acids are affected by processing methods. Hyperspectral imaging is a novel technique widely used for food quality and safety evaluation. In the current study, the contents of monounsaturated and polyunsaturated fatty acids were assessed by Hyperspectral imaging. Optimal wavelengths were selected by the regression coefficients curves of partial least squares regression models. The least-squares support vector machine models established achieved a better coefficient of determination in the Monte Carlo validation set than the partial least squares regression models developed and the R²MV values for the least squares - support vector machine models based on selected optimal wavelengths were higher than 0.81. Finally, colour maps of the contents of monounsaturated and polyunsaturated fatty acids were developed.
Firmness is an important quality indicator of fruit and is closely related to physical structures and mechanical properties. In this study, an online detection system using a laser Doppler vibrometer (LDV) was developed to acquire the acoustic vibration response signals of ‘Cuiguan’ pears for firmness detection. Based on compositional analysis of pear tissue, the predominant components in the process of fruit softening were crude fiber and pectin. Subsequently, different classification models for prediction of pear firmness were established based on sensory evaluation. The results showed that a back propagation neural network (BPNN) method using the elasticity index (EI), peak value at f2 (A) and peak area (S) as input variables had the highest discriminant accuracy. The accuracies of the calibration and validation sets were 93.3 % and 90.5 %, respectively. Moreover, the stiffness in the peeled group obtained by a puncture test was regarded as a dependent variable in quantitative analysis due to its high correlation coefficients with sensory scores and chemical indices in postharvest. In addition, multiple regression models were compared with simple linear regression models. The highest correlation coefficient rp of the prediction set was observed for the BPNN model. In addition, the BPNN method using EI, A, S and the shape index had the best prediction performance. The correlation coefficient rp and RMSEP of the prediction set were 0.832 and 0.277 N mm⁻¹, respectively.
The challenge of deriving quantitative information from the infrared spectra of proteins arises from the large number of secondary structures and amino acid side-chain functional groups that all contribute to the spectral intensity, such as within the amide I band (1600-1700 cm-1). The band is invariably heavily convoluted from overlapping spectral features, thereby making interpretation difficult such that deconvolution is usually required. This work critically examines the methods available to deconvolute the spectra and assesses the commonly used methods and algorithms applied to vibrational spectra for smoothing and peak identification. We show that unless their spectra have very high signal-to-noise ratios, quantitative analysis to decipher protein constituents is not feasible. The advantages and disadvantages of spectral smoothing using adjacent averaging, the Savitzky-Golay filter and the fast Fourier transform filter are examined in detail. The use of derivative spectra to identify peaks is described with particular reference to the influence and reduction of interfering water bands in the amide I region. The reliability of band narrowing techniques such as second-derivative analysis or Fourier deconvolution that lead to the identification of the contributing protein peaks is investigated. Both methods are shown to be limited in their capacity to resolve features with very similar frequencies. Additionally, the presence of narrow bands arising from high-frequency noise whether from atmospheric water vapor, acoustic vibrations, or electrical interference results in both methods becoming increasingly unusable as narrow bands are preferentially enhanced at the expense of broad ones such as the amide I bands. An optimal strategy is critically developed to allow accurate determination and quantification of protein constituents and their conformations. Additionally, quantitative methods are proposed to account for baseline shifts, which would otherwise introduce significant errors in similarity indices.
Near-infrared (NIR) and mid-infrared (MIR) hyperspectral techniques in tandem with chemometric analyses were employed for developing multispectral real-time systems allowing the food industry to monitor moisture content (MC) in tubers including various potato and sweet potato products during drying. Multivariate models were established by partial least-squares regression (PLSR), support vector machine regression (SVMR), locally weighted partial least square regression (LWPLSR), and back propagation artificial neural network (BPANN) using full spectral ranges of 10372–6105 cm-1 (Spectral Set Ⅰ), 3996–600 cm-1 (Spectral Set Ⅱ), and 1700–900 cm-1 (Spectral Set Ⅲ). The LWPLSR from Spectral Set Ⅰ and BPANN from Spectral Set Ⅱ and Ⅲ, obtained the highest accuracies for tuber MC prediction. Then, both regression coefficient (RC) and successive projection algorithm (SPA) were respectively used for the selection of feature wavelengths in Spectral Set Ⅰ, Ⅱ and Ⅲ. Instead of choosing many groups of characteristic variables for different varieties of potatoes and sweet potatoes, one set of feature variables for all tubers was selected from each spectral region for the convenience of industrial application. Eventually, six sets of feature wavelengths chosen from Spectral Set Ⅰ, Ⅱ and Ⅲ were used to optimize models. The simplified SPA-LWPLSR from Spectral Set Ⅱ and SPA-BPANN from Spectral Set Ⅲ acquired good model performances for the tuber MC prediction, with determination coefficients in prediction (R2P) of 0.950 and 0.904, respectively. The RC-BPANN model from Spectral Set Ⅰ achieved the highest R2P of 0.965. Such accuracies were comparable to that of full spectral models. The results reveal that hyperspectral techniques have great potential in the food industry for real-time measurement of tuber MC.
The potential of chemical imaging for rapid measurement of dry matter concentration (DMC) and starch concentration (SC) in both potato and sweet potato tubers was investigated. The time series images of tuber samples were acquired, then the resulting reflectance spectra (RS) were corrected and transformed into absorbance spectra (AS), and exponent spectra (ES). Full wavelength regression models including multiple linear regression (MLR), partial least squares regression (PLSR) and locally weighted partial least squares regression (LWPLSR) were established based on spectral profiles with measured DMC and SC values. The best calibration model for measuring DMC and SC was LWPLSR based on ES and RS where the coefficients of determination in cross-validation (R2CV) were 0.987 and 0.985, and the root mean squared errors in cross-validation (RMSECV) were 0.015 and 0.014, respectively. After, six groups of eight feature wavelengths were chosen from RS, AS and ES based on wavelength selection methods including β-coefficient (βC) of PLSR and the first derivative and mean centering iteration algorithm (FMCIA), and were successively used to build simplified models. The acquired FMCIA-RS-LWPLSR and βC-RS-LWPLSR models showed better accuracy than other simplified models, with R2P of 0.985 and RMSEP of 0.016 for DMC prediction, and R2P of 0.983 and RMSEP of 0.015 for SC prediction, respectively. Besides, the optimal models for MLR and PLSR were obtained using FMCIA on the basis of the ES. After further reducing the number of feature wavelengths, only six wavelengths (1028, 1068, 1135, 1208, 1262 and 1460 nm) were selected and utilized to develop the simplest FMCIA-Es-MLR model for determining DMC and FMCIA-Es-PLSR model for detecting SC, yielding a reasonable level of accuracy with R2P of 0.962 and 0.963 as well as RMSEP of 0.025 and 0.023, respectively. Furthermore, the time series variations of DMC and SC on tuber samples were visualized based on an equation to apply the simplest models to the spectral images.
Image analysis involving mathematical, statistical and software programming approaches are the essential elements of any computer-integrated hyperspectral imaging systems. The theoretical and practical issues associated with the development, analysis, and application of essential image processing algorithms are explored in order to exploit hyperspectral imaging for application to food quality evaluations. The breadth of different image processing approaches adopted over the years in attempting to implement hyperspectral imaging for food quality monitoring was surveyed. Firstly, the fundamental configurations and working principles of hyperspectral systems, as well as the basic concept and structure of hyperspectral data, were described and explained. The understanding of different approaches used during image acquisition, data collection and visualisation were examined. Strategies and essential image processing routines necessary for making the appropriate decision during detection, classification, identification, quantification and/or prediction processes are presented. Examples and figures were selected to reinforce the main approach of each analysis algorithm applied in different agro-food products to answer the question “What does the user want to see in the target food samples?” The theoretical background for each algorithm was beyond the scope of this article thus only essential equations were addressed. The literature presented clearly revealed that hyperspectral imaging systems have gained a rapid interest from researchers to display the chemical structure and related physical properties of numerous types of food stuffs and hyperspectral imaging systems are expected to gain more considerably more potential and application in food processing and engineering plants.
Hyperspectral imaging techniques have widely demonstrated their usefulness in different areas of interest in pharmaceutical research during the last decade. In particular, middle infrared, near infrared, and Raman methods have gained special relevance. This rapid increase has been promoted by the capability of hyperspectral techniques to provide robust and reliable chemical and spatial information on the distribution of components in pharmaceutical solid dosage forms. Furthermore, the valuable combination of hyperspectral imaging devices with adequate data processing techniques offers the perfect landscape for developing new methods for scanning and analyzing surfaces. Nevertheless, the instrumentation and subsequent data analysis are not exempt from issues that must be thoughtfully considered. This paper describes and discusses the main advantages and drawbacks of the measurements and data analysis of hyperspectral imaging techniques in the development of solid dosage forms.
SOP for Determination of Dry Matter Content
- M Adesokan
- E Alamu
- B Maziya-Dixon
Adesokan M., Alamu E. and Maziya-Dixon B., SOP for Determination of Dry Matter
Content. RTBfoods Project Report, Ibadan, Nigeria, p. 7 (2020) 〈https://mel.cgiar.
org/reporting/download/report_file_id/17813〉. (Accessed on February 16, 2024).
Prediction of yam cooking behaviour using hyperspectral imaging
- Meghar
Application of Hyperspectral Technology in Potato Variety Identification and Quality Non-destructive Testing
- W Jiang
Jiang, W., 2017. Application of Hyperspectral Technology in Potato Variety
Identification and Quality Non-destructive Testing. Northeast Agricultural
University, Ph.D. thesis, China.
Prediction of yam cooking behaviour using hyperspectral imaging. Rep. HSI calibrations Dry. Matter, pectin, starch Texture raw fresh yam slices CIRAD Fr
- K Meghar
- J Boyer
- F Davrieux
Meghar, K., Boyer, J., Davrieux, F., 2022. Prediction of yam cooking behaviour using
hyperspectral imaging. Rep. HSI calibrations Dry. Matter, pectin, starch Texture raw
fresh yam slices CIRAD Fr. https://doi.org/10.18167/agritrop/00707.