Hyperspectral image classification of wolfberry with different geographical origins based on three-dimensional convolutional neural network

Mu, Qingshuang; Zhang, Kang; Guo, Yanju; Chen, Lei; Wang, Shenyi; Zhao, Yuchen

doi:10.1080/10942912.2021.1987457

Cited by 15 publications

(8 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These results are similar to other studies on the original classification of wolfberries [2][3][4][5][6][7], but the previous studies on the classification of wolfberries were of different varieties. For example, LS-SVM was used by Li et al [2] to calibrate the discriminative model of superior quality and inferior quality black wolfberries in Luo Mu Hong, Qinghai-Tibet Plateau, and Xinjiang; Shen et al [4] used near-infrared spectroscopy and chemometrics to determine the geographic origin of wolfberries in the North China Plain, Loess Plateau, Northeast China Plain, and the Northwest Basin; Mu et al [6] classified HSI images of wolfberries from four origins in Ningxia, Qinghai, Xinjiang, and Gansu. These origins have relatively far distances, large differences in water and soil environment and growth conditions, and so can be identified easily.…”

Section: Comparison Of Multi-task Datasetssupporting

confidence: 91%

“…After removing the bands with noise at the front and back, it was observed that the curve begins to diverge after the first absorption peak at 1040 nm, which is due to the second vibration of the N-H bond in the protein or amino acid. There were significant differences in the spectral curves from 1100 nm to 1400 nm, which are close to the double-frequency absorption band of the C-H bond [6]. The absorption peak is due to the secondary stretching vibration of the C-H bond in the protein, starch, or lipid.…”

Section: Overview Of Spectral Profilesmentioning

confidence: 83%

“…They include isotope ratio mass spectrometry (IRMS), inductively coupled plasma mass spectrometry (ICP-MS), high-performance liquid chromatography (HPLC), gas chromatography (GC), and liquid chromatography-mass spectrometry (LC-MS/MS). These methods require skilled personnel to operate and are not only costly but also destructive to samples [5][6][7]. There are no significant differences in the levels of trace elements among different origins, so these techniques, such as ICP-MS, have limitations in the identification of the near geographical location of the fruit.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

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

Section: Comparison Of Multi-task Datasetssupporting

confidence: 91%

Section: Overview Of Spectral Profilesmentioning

confidence: 83%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

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

show abstract

“…The peak region near 1,465 nm was related to the firstorder frequency doubling of the O-H bond stretching vibration (Aernouts et al, 2011). The characteristics of the Lycium barbarum spectral curve found in this study were consistent with the results of Mu et al (2021) for the dried fruits of wolfberries from Ningxia, Qinghai, Gansu, and Xinjiang of China. In addition, similar conclusions were reported by Zhang et al (2020) who used hyperspectral analysis and predicted the contents of total phenols, total flavonoids, and total anthocyanins in the dried fruits of black goji berry (Lycium ruthenicum Murr.)…”

Section: Hyperspectral Characteristics and Transformationsupporting

confidence: 88%

Estimation of nutrient contents in wolfberry (Lycium barbarum L.) based on hyperspectral analysis

ZHAO

ZHANG

Zhang

et al. 2022

Not Bot Horti Agrobo

View full text Add to dashboard Cite

Rapid and accurate determination of the nutrient contents in wolfberry (Lycium barbarum L.) is of great significance for identifying the quality and origin of the fruit. Compared to traditional chemical analysis methods, hyperspectral remote sensing has the advantages of high speed and low cost. In this study, the dried fruits of wolfberry (cultivar ‘Ningqi No. 7’) taken from the Huinong, Yinchuan, Zhongning, and Tongxin regions of Ningxia, China, in 2020 were selected as samples. Two methods, the variable importance measure (VIM) under random forest and the successive projection algorithm (SPA), were applied to select the hyperspectral characteristic variables. The test set coefficient of determination (R2p), root mean square error of prediction (RMSEP), and relative percent deviation (RPD) for different pre-treatments and characteristic variable selection methods were compared. Finally, the optimal estimation models of partial least squares regression (PLSR) for the contents of eight nutrients in wolfberry were established. The results were as follows: (1) The variation trends of the original hyperspectral reflectance curves of wolfberries from different yield areas and harvest dates were similar, and the spectral characteristic absorption bands were significantly enhanced after first-order differential transformation. (2) After extracting the characteristic bands using SPA, the RPD values were all above 2.0, and the estimation performance was significantly better than that of the full bands. (3) The optimal estimation model for total sugar (TS), crude protein (CP), and Ca was found to be SG-MSC-SPA-PLSR; the optimal estimation model for Lycium barbarum polysaccharide (LBP), Mn, and Zn was found to be SG-2ndD-SPA-PLSR; and the optimal estimation model for Cu and Fe was found to be SG-1stD-SPA-PLSR. The results have important reference value for the quality evaluation and origin traceability of wolfberry.

show abstract

“…Tang et al (2021) obtained 88.33% prediction accuracy from HSIs of six varieties of L. barbarum grains using competitive adaptive reweighted sampling (CARS) algorithms for wavelength selection, the whale optimization algorithm for model enhancement, and the SVM as the prediction model. Mu et al (2021) obtained a 99% classification accuracy of wolfberry fruit from different origins using HSI and a hybrid‐convolutional neural network. Rice seeds were also identified using a combination of HSI technology and CNN with an 85% model performance (Qiu et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

Nondestructive detection of adulterated wolfberry (Lycium Chinense) fruits based on hyperspectral imaging technology

Nirere

Sun

Kama

et al. 2023

J Food Process Engineering

View full text Add to dashboard Cite

In order to detect adulterants on Lycium Chinense species effectively, a rapid, clean, and nondestructive detection method based on hyperspectral imaging (HSI) technology was conducted in a wavelength range of 400.68-1001.60 nm. Industrial sulfur particles were chosen as a dye to prepare three groups of adulterated L. Chinense samples as the research object. The whole L. Chinense was considered the region of interest. First, a multiple scatter correction (MSC) method was used to preprocess spectra data. The competitive adaptive reweighted sampling (CARS) and linear discriminant analysis approaches were contrasted for optimal extraction of wavelengths characteristic. Then, two models were established: K-nearest neighbor (KNN) and support vector machine (SVM). Furthermore, the performance accuracies of KNN and SVM models were compared. According to the outcomes, the SVM model built on CARS provided the best classification impact. The accuracy for the prediction set was 98.75%, and the accuracy for the training set was 100%. Also, the kernel parameters c and g of the SVM model were enhanced by genetic algorithm (GA) optimization. The values parameters (c, g) were set at 14.975 and 0.224, respectively, and the results improved by 1.25% at an elapsed time of 1.887 s, with the accuracy reaching 100% on both training and test sets. This study aimed to detect and classify sulfur-adulterated wolfberries using an improved SVM and HSI. Finally, the results demonstrate that a combination of HSI and the CARS-GA-SVM model could be used for the rapid detection of foreign entities' in wolfberry fruits. Practical applicationsDried wolfberry adulteration has a direct link to the overall quality of the fruits, and potentially compromises the health of the fruits consumers. The traditional methods of testing adulterants on Lycium Chinense are arduous, require a lot of time, and are highly impacted by biased elements, necessitating new techniques. HSI technology, on the other hand, is nondestructive, quick/fast, accurate, subjective, reproducible, and pollution-free. The study findings proved to be recommendable for initiating a feasible mobile system for rapidly detecting adulteration on L. Chinense.

show abstract

Hyperspectral image classification of wolfberry with different geographical origins based on three-dimensional convolutional neural network

Cited by 15 publications

References 35 publications

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

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

Estimation of nutrient contents in wolfberry (Lycium barbarum L.) based on hyperspectral analysis

Nondestructive detection of adulterated wolfberry (Lycium Chinense) fruits based on hyperspectral imaging technology

Contact Info

Product

Resources

About