Application of Convolutional Neural Network-Based Feature Extraction and Data Fusion for Geographical Origin Identification of Radix Astragali by Visible/Short-Wave Near-Infrared and Near Infrared Hyperspectral Imaging

Xiao, Qinlin; Bai, Xiulin; Gao, Pan; He, Yong

doi:10.3390/s20174940

Cited by 28 publications

(15 citation statements)

References 38 publications

(36 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared with the best model with a single Vis/NIR, the overall classification accuracy of the prediction set is improved from 78% to 92%. Xiao et al (2020) use PCA to extract the spectral features of Vis (380–1030 nm) and NIR (874–1734 nm), and fused them to develop a logistic regression model in order to identify the origin of Radix Astragali. Compared with the optimal model of a single Vis/NIR, the overall classification accuracy of the prediction set is improved from 99.453% to 99.767%.…”

Section: Resultsmentioning

confidence: 99%

Rapid and accurate detection of starch content in mixed sorghum by hyperspectral imaging combined with data fusion technology

Jiang

Tian

et al. 2022

J Food Process Engineering

View full text Add to dashboard Cite

Sorghum is a type of brewing material. The starch content of different kinds of mixed sorghum affect the quality and yield of liquor. Therefore, an accurate and efficient detection of its starch content is of great significance to obtain high‐quality and high‐yield of liquor. Based on the data of single visible light (Vis), near‐infrared (NIR), and Vis and NIR fusion, this study develops a genetic algorithm optimization BP neural network (GA‐BPNN) model and a particle swarm optimization support vector machine model. It is deduced that the model based on the data from the Vis and NIR fusion has the highest performance in predicting the starch content of mixed sorghum. For the prediction of amylose, the GA‐BPNN model developed after using the Pearson correlation coefficient to select the characteristic wavelength for fusion, has the highest performance (RMSEP = .0298 and = .9948). For the prediction of amylopectin, the GA‐BPNN model developed after fusion of spectral features extracted by principal component analysis, has the highest performance (RMSEP = .0213 and = .9985). In conclusion, the hyperspectral imaging combined with data fusion can rapidly and accurately detect the starch content of mixed sorghum.Practical ApplicationsAs the single raw material of Maotai‐flavor liquor and the main raw material of Luzhou‐flavor liquor, the amylopectin content of sorghum directly affects the quality of liquor. Its amylose content also affects the yield of liquor. Because the different varieties of sorghum have different amylose and amylopectin content, liquor manufacturers often use several mixed varieties of sorghum as a brewing material. In order to ensure the quality and yield of liquor, it is particularly crucial to detect the amylose and amylopectin content variables of blended sorghum. The traditional sorghum starch content determination methods, such as the chemical detection methods and near infrared spectroscopy, have the disadvantages of destructiveness, low efficiency, and low detection accuracy. In the process of liquor brewing, the traditional detection methods cannot guide the timely adjustment of brewing process parameters. The hyperspectral imaging technology is widely used in food material detection, due to its fast and nondestructive advantages. The data fusion technology can fuse data from different sources of the same object to be detected, in order to obtain more comprehensive data for the development of accurate prediction models. This study uses the hyperspectral imaging combined with data fusion in order to quickly and accurately predicts the starch content (amylose, amylopectin) of mixed sorghum, which has a guiding significance for the timely adjustment of process parameters in the brewing process of liquor. In addition, it provides an accurate method for the component detection of other grains.

show abstract

Section: Resultsmentioning

confidence: 99%

Rapid and accurate detection of starch content in mixed sorghum by hyperspectral imaging combined with data fusion technology

Jiang

Tian

et al. 2022

J Food Process Engineering

View full text Add to dashboard Cite

show abstract

“…Astragali Radix (AR) can tonify the middle-jiao and replenish qi, solidify the surface and promote diuresis, dispel sepsis and renew muscle [6]. Modern research shows that AR has many biological functions, such as vasodilation [7], antioxidant activity [8,9], immunomodulation [10,11], antiaging activity [12,13], and antitumor activity [14]. The process of producing AR decoction pieces is "remove impurities, separate size, cleanse, moisten, cut thick pieces and dry [15]."…”

Section: Discussionmentioning

confidence: 99%

Research on the drying Mechanism of Astragali Radix during Processing

Peng¹,

Wu²,

Ye³

et al. 2021

Preprint

View full text Add to dashboard Cite

Backgroundstragali Radix (Huangqi) is an important Chinese herbal medicine that is always processed into pieces for clinical use. There are many operations required after harvesting, among which drying of AR pieces is a key step. Unfortunately, research on the drying mechanism is still limited.MethodsLow-field nuclear magnetic resonance (LF-NMR) and magnetic resonance imaging (MRI) techniques were applied to study the water status and water distribution during the drying process. The contents of bioactive ingredients and texture changes were measured by HPLC and texture analyzer, respectively.ResultsDuring the drying process, the moisture content first decreased rapidly. After a particular time, the moisture content decreased slowly, and the equilibrium value was reached. The trends for changes in flavonoid glycoside and flavonoid aglycone contents in AR pieces were entirely opposite. The relaxation times T22 and T23 and the peak areas A22 and ATotal decreased significantly during drying. In addition, the intensity of the water signal decreased from the surface to the interior. The textural parameters hardness, adhesion and fracturability increased without an apparent change in springiness. The parameters of LF-NMR and the texture analyzer were correlated with water content.ConclusionsThis study indicated that HPLC, LF-NMR, MRI and texture analyzers provide a scientific basis for elucidating the principles for drying of AR pieces. The method is useful and shows potential for popularization and application; accordingly, it can easily be extended to other radix and/or rhizoma Traditional Chinese Medicinal materials.

show abstract

“…In the deep learning model, gradient descent optimizes the weight coefficients and deviations to keep the loss function as small as possible [21]. The learning rate is the most important hyperparameter [8,9]. It directly affects the gradient convergence rate and iteration times.…”

Section: Algorithm Optimizationmentioning

confidence: 99%

“…Hyperspectral imaging (HSI) technology and machine vision have shown great potential in the field of adulteration identification. Xiao et al [8] applied HSI to identify Radix Astragali from five different geographical origins. Studies by Fazari et al [9] and Zhang et al [10] found that image information from HSI enhanced model performance.…”

Section: Introductionmentioning

confidence: 99%

Identification of Near Geographical Origin of Wolfberries by a Combination of Hyperspectral Imaging and Multi-Task Residual Fully Convolutional Network

Cui

Hao

et al. 2022

Foods

View full text Add to dashboard Cite

Ningxia wolfberry is the only wolfberry product with medicinal value in China. However, the nutritional elements, active ingredients, and economic value of the wolfberry vary considerably among different origins in Ningxia. It is difficult to determine the origin of wolfberry by traditional methods due to the same variety, similar origins, and external characteristics. In the study, we have for the first time used a multi-task residual fully convolutional network (MRes-FCN) under Bayesian optimized architecture for imaging from visible-near-infrared (Vis-NIR, 400–1000 nm) and near-infrared (NIR-1700 nm) hyperspectral imaging (HSI) technology to establish a classification model for near geographic origin of Ningxia wolfberries (Zhongning, Guyuan, Tongxin, and Huinong). The denoising auto-encoder (DAE) was used to generate augmented data, then principal component analysis (PCA) was combined with gray level co-occurrence matrix (GLCM) to extract the texture features. Finally, three datasets (HSI, DAE, and texture) were added to the multi-task model. The reshaped data were up-sampled using transposed convolution. After data-sparse processing, the backbone network was imported to train the model. The results showed that the MRes-FCN model exhibited excellent performance, with the accuracies of the full spectrum and optimum characteristic spectrum of 95.54% and 96.43%, respectively. This study has demonstrated that the MRes-FCN model based on Bayesian optimization and DAE data augmentation strategy may be used to identify the near geographical origin of wolfberries.

show abstract

Application of Convolutional Neural Network-Based Feature Extraction and Data Fusion for Geographical Origin Identification of Radix Astragali by Visible/Short-Wave Near-Infrared and Near Infrared Hyperspectral Imaging

Cited by 28 publications

References 38 publications

Rapid and accurate detection of starch content in mixed sorghum by hyperspectral imaging combined with data fusion technology

Rapid and accurate detection of starch content in mixed sorghum by hyperspectral imaging combined with data fusion technology

Research on the drying Mechanism of Astragali Radix during Processing

Identification of Near Geographical Origin of Wolfberries by a Combination of Hyperspectral Imaging and Multi-Task Residual Fully Convolutional Network

Contact Info

Product

Resources

About