Search

The moisture content (MC) and acidity content (AC) of the fermented grains used in liquor production directly affect the liquor quality and yield; as such, they are important indicators used to evaluate the quality of fermented grains. In this study, extreme gradient enhancement algorithm (XGBoost), partial least square regression (PLSR), and extreme learning machine (ELM) models were developed based on spectral data collected by near-infrared (NIR) hyperspectral imaging (HSI) technology. First, PLSR models were established after SNV and MSC algorithms preprocessed the HSI data, and the best preprocessing method was determined (MC: SNV; AC: MSC). Then, the competitive adaptive reweighting sampling (CARS) algorithm and principal component analysis (PCA), both combined with the successive projection algorithm (SPA), were used to extract the characteristic wavelengths from the full-band spectral data. Ultimately, the XGBoost model developed using the characteristic wavelengths extracted by CARS-SPA most accurately predicted the MC (RPD = 6.4167, RP2 = 0.9757, RMSEP = 0.0442 g·100 g−1) and AC (RPD = 13.0308, RP2 = 0.9941, RMSEP = 0.0216 mmol·10 g−1). The results showed that the XGBoost model could more accurately predict the MC and AC of the fermented grains from hyperspectral images of the grains, providing an effective method for the rapid analysis of raw materials used in the fermentation of liquor. [ABSTRACT FROM AUTHOR]

Starch and alcohol serve as pivotal indicators in assessing the quality of lees fermentation. In this paper, two hyperspectral imaging (HSI) techniques (visible–near‐infrared (Vis–NIR) and NIR) were utilized to acquire separate HSI data, which were then fused and analyzed toforecast the starch and alcohol contents during the fermentation of lees. Five preprocessing methods were first used to preprocess the Vis–NIR, NIR, and the fused Vis–NIR and NIR data, after which partial least squares regression models were established to determine the best preprocessing method. Following, competitive adaptive reweighted sampling, successive projection algorithm, and principal component analysis algorithms were used to extract the characteristic wavelengths to accurately predict the starch and alcohol levels. Finally, support vector machine (SVM)‐AdaBoost and XGBoost models were built based on the low‐level fusion (LLF) and intermediate‐level fusion (ILF) of single Vis–NIR and NIR as well as the fused data. The results showed that the SVM‐AdaBoost model built using the LLF data afterpreprocessing by standard normalized variable was most accurate for predicting the starch content, with an RP2$\ R_P^2$ of 0.9976 and a root mean square error of prediction (RMSEP) of 0.0992. The XGBoost model built using ILF data was most accurate for predicting the alcohol content, with an RP2$R_P^2$ of 0.9969 and an RMSEP of 0.0605. In conclusion, the analysis of fused data from distinct HSI technologies facilitates rapid and precise determination of the starch and alcohol contents in fermented grains. [ABSTRACT FROM AUTHOR]

The total acid content of Daqu during fermentation directly influences the quality of the Daqu. Hyperspectral imaging (HSI) was used to rapidly and accurately detect the total acid content of Daqu. The performances of two spectral ranges obtained by two hyperspectral cameras for Daqu total acid content were compared. Different pretreatment methods were used to build the partial least squares regression (PLSR) model, and the optimal pretreatment for the corresponding spectra was determined. A combination of principal component analysis (PCA) and successive projection algorithm (SPA) was used to select the feature wavelengths. The PCA algorithm optimized by SPA was more accurate than the PCA algorithm alone when extracting the feature wavelengths. PLSR, backpropagation neural network (BP), and the backpropagation‐genetic algorithm (BP‐GA) model were developed to predict the total acid content of the Daqu samples by using the full wavelengths and feature wavelengths, respectively. The predictive performance of each model was compared and analyzed. The optimal model was the MSC + PCA‐SPA+BP‐GA model (Rc2 = 0.8832, RMSEC = 0.1163, Rp2 = 0.9724, RMSEP = 0.0593, RPD = 4.2860, AB_RMSE = 0.0570). In addition, the distribution of total acid content within the region of interest (ROI) of Daqu samples was visualized at different fermentation stages using the optimized model. These results show that HSI technology can rapidly and precisely detect the total acid content of Daqu and visualize the content distribution, which provides important technical support for upgrading and transforming the liquor. Practical Applications: Daqu is the saccharifying agent, leavening agent, and aroma‐generating agent of liquor brewing. During the fermentation process, the total acid content is an important indicator for the quality evaluation of Daqu. Therefore, it is highly significant to detect the total acid content in Daqu quickly and accurately. The HSI technique is widely used to detect the contents of various substances in food. In this study, the HSI technique combining the PCA‐SPA and BP‐GA algorithms was established to detect the total acid content of Daqu, and a satisfactory result was obtained, showing that HSI technology combined with an optimized algorithm could be used to realize rapid and accurate detection of the total acid content of Daqu. In addition, the predictive model proposed in this study is not only applicable for detecting other food substances but also this method has great potential for real‐time indicators detection in winery for future work. [ABSTRACT FROM AUTHOR]

The rapid differentiation of the type and proportion of raw materials is the key factor for the adjustment of the subsequent process parameters and to ensure product quality. Therefore, in this study, a combination of an optimization algorithm and hyperspectral imaging (HSI) technology was used to distinguish many varieties of broken grains and determine their mixing ratio. The reflectance of the region of interest (ROI) of the sample was extracted, and the outliers were then removed by the density-based spatial clustering of application with noise (DBSCAN) method and the Mahalanobis distance (MD). The principal component analysis (PCA) algorithm was then used to analyze the spectral data after removing the outliers. Next, iterative variable subset optimization (IVSO) and competitive adaptive reweighting sampling (CARS) were used to extract the feature wavelengths. Four classification and recognition models (PNN, GRNN, BPNN, and RBFNN) based on the full and characteristic wavelengths were then established. A comparison of the results revealed that the BPNN model achieved the best effect; the classification recognition accuracies of the training set, testing set, and verification set of the full and characteristic wavelength models were above 99%. To determine the differences between BPNN models based on the full and characteristic wavelengths, seven different kernel functions were used to train the BPNN models; the classification recognition accuracy of the BPNN models trained by the one-step secant (OSS) and resilient functions was greater than 99%. Therefore, by using a combination of an optimization algorithm and HSI technology, it is possible to quickly distinguish mixed grains and determine their mixing ratio. This study provides new technical guidance for the liquor industry to improve the quality of liquor. [Display omitted] • The Otsu method combined with watershed algorithm has a high success rate in extracting spectral reflectance of samples. • The IVSO-CARS algorithm is a new and efficient method for feature wavelength screening. • The classification and recognition accuracy of BPNN model can reach 99%. [ABSTRACT FROM AUTHOR]

• A robust model was developed and validated for evaluation of sorghum quality. • Isolated forest and CARS-SPA were conducive to the optimization of the model. • The average CRR of the model in identifying sorghum varieties was greater than 91%. • The deviation of the predicted mixing ratio of adulterated samples was less than 4%. This study combined hyperspectral imaging (HSI) and deep forest (DF) to develop a reliable model for conducting a rapid and nondestructive determination of sorghum purity. Isolated forest (IF) algorithm and principal component analysis (PCA) were used to remove the abnormal data of sorghum grains. Competitive adaptive reweighted sampling (CARS) algorithm and successive projections algorithm (SPA) were combined and used to extract the characteristic wavelengths. Gray-level co-occurrence matrix (GLCM) was used to extract the textural features. DF models were established based on the different types of data. Specifically, the DF models established using the characteristic spectra produced the best recognition results: the average correct recognition rate (CRR) of the models was greater than 91%. In addition, the average CRR of validation set Ⅰ was 88.89%. These results show that a combination of HSI and DF could be used for the rapid and nondestructive determination of sorghum purity. [ABSTRACT FROM AUTHOR]

• CF was used to predict the contents of amylose and amylopectin in sorghum. • PCA-SPA was used to extract characteristic wavelengths corresponding to amylose and amylopectin. • The average RPD values of the CF models were both greater than 4.7. • Accurate recognition for the varieties of sorghum grains was realized. • This study provides a new method for the starch content detection of cereals. The contents of amylose and amylopectin in sorghum directly affects the quality and yield of liquor. Hyperspectral imaging (HSI) is an emerging technology widely applied in the content analysis of food ingredients. In this study, the effects of different preprocessing methods on visible-light and near-infrared spectral data were analyzed, and the prediction accuracies of these spectral data were compared. Principal components analysis (PCA) and successive projections algorithm (SPA) were combined to extract the characteristic wavelengths. Using both the full and characteristic wavelengths, partial least square regression (PLSR) and cascade forest (CF) models were developed to predict the contents of amylose and amylopectin in different varieties of sorghum. The average RPD values of the CF models established by the characteristic wavelengths were 4.7622 and 5.5889, respectively. These results corroborated the utility of HSI in achieving the rapid and nondestructive prediction of amylose and amylopectin contents in different varieties of sorghum. [ABSTRACT FROM AUTHOR]

• The R P 2 values of the CF models were greater than 98.00 %. • PCA-SPA was used to extract characteristic wavelengths. • The distribution map of the substance content was obtained via visualization. • This study provides a new method for obtaining real-time index data on Daqu. In this research, the hyperspectral imaging technique was employed to realize the rapid and accurate detection of the reducing sugar and amino acid nitrogen contents of Daqu during fermentation. Different preprocessed data were employed to establish cascade forest (CF) models to determine the best preprocessing method. Principal component analysis (PCA) and the successive projection algorithm (SPA) were then combined to extract the characteristic wavelengths. Four types of models (CF, BPNN, SVR, and PLSR) were established based on the full and characteristic wavelengths, respectively. Among all the models, the CF model established by the characteristic wavelengths was determined to be superior for the detection of the reducing sugar and amino acid nitrogen contents: for reducing sugar, R P 2 = 0.9862, RMESP = 0.0812 g/100 g, and RPD = 6.0402; for amino acid nitrogen, R P 2 = 0.9876, RMSEP = 0.0500 g/100 g, and RPD = 6.3698. The best model achieved the visualization of the reducing sugar and amino acid nitrogen contents in the region of interest. The satisfactory detection results demonstrate that the hyperspectral imaging technique can be used to realize the rapid and accurate detection of the reducing sugar and amino acid nitrogen contents of Daqu during fermentation. [ABSTRACT FROM AUTHOR]

• First time HSI was used to detect adulterated sorghum. • Watershed algorithm could effectively segment sorghum grains. • PCA and CA could effectively find abnormal samples in a single sorghum variety. • PLS-DA model could rapidly and nondestructively detect sorghum adulteration. • The distribution map of the sorghum was obtained through visualization. This paper proposes a sorghum adulteration detection model using hyperspectral imaging technology (HSI), image processing technology, and multivariate analysis technology. The model used a watershed algorithm to extract hyperspectral data from sorghum grains. Principal component analysis (PCA) and clustering analysis (CA) were used to remove abnormal samples of sorghum. Partial least squares discriminant analysis (PLS-DA) was used to identify the variety of sample, and a sorghum distribution map and adulteration ratios were obtained by marking varieties with different colors. This paper presents, for the first time, HSI use for identification of adulteration in sorghum using PCA and CA. Accuracy of the model identification for the validation set reached 96%, and for the adulterated samples reached 91%, and comprehensive accuracy of the model could reach more than 90%. These results show that the model can rapidly and nondestructively detect sorghum adulteration. [ABSTRACT FROM AUTHOR]

Rice is one of the main raw materials used in the brewing of Chinese liquor. The fatty acid content of rice is one of the most important indicators of the degree of aging of rice, which greatly influences the flavor of the liquor. In this study, hyperspectral imaging (HSI) was used in combination with an integrated learning model to predict the fatty acid content in rice. Spectral, textural, and color information were extracted from the hyperspectral images of the rice grains derived from at different levels of aging. After analyzing, the best preprocessing method was determined to be SG-SNV. Principal component analysis with different kernel functions (KPCA) was used to extract the characteristic wavelengths, and PSO-SVR, SSA-KELM, and SVM-AdaBoost models were developed to predict the fatty acid content of rice based on full-wavelength spectral data, feature wavelengths, feature wavelengths combined with texture information, and feature wavelengths combined with color information. Comparison of model performance showed that the SVM-AdaBoost model, which was built by combining the texture information of the rice grains with the feature wavelengths, had the highest performance of all the models, with a RPD value of 22.3663. These results showed that combining HSI with an integrated learning model can be effective for the rapid and accurate prediction of fatty acid content in rice. [Display omitted] [ABSTRACT FROM AUTHOR]