ArticlePDF Available

Hyperspectral Detection of Sugar Content for Sugar-sweetened Apples Based on Sample Grouping and SPA Feature Selecting Methods

Authors:
Article

Hyperspectral Detection of Sugar Content for Sugar-sweetened Apples Based on Sample Grouping and SPA Feature Selecting Methods

Abstract

This paper aims to select the spectral feature grouping method and establish regression detection models based on feature grouping and successive projection algorithms (SPA) feature optimization. At first, the hyperspectral images and sugar content of 174 apple samples were collected to extract the hyperspectral data using ENVI 4.7 software. After pretreating the spectral data by using multiple scattering correction (MSC), all samples were stored according to sugar content. In the case that the training set and the test set were divided into different groups at the same allocation ratio, the full-band grouping modeling method and the continuous projection extraction feature wavelength grouping modeling method were used to model the hyperspectral data and the sugar content data, respectively. By comparing the prediction effect in terms of full-band group modeling in which we character the grouping method as 1, 2, and 3, respectively. In the case of grouping method 1, i.e., when the test set and training set were randomly assigned according to the set ratio column, the prediction effect of the model established was optimal. Nonetheless, the simulation result was stable because it was the averaging R of 100 running results. In the case of grouping method 2, i.e., when the training set and the test set were divided using interval allocation according to the set ratio, the prediction effect of the model, was slightly worse than that of grouping method 1. Grouping method 3, i.e., the training set and the test set were grouped in series according to the set ratio, has the worst prediction effect of the model. Results indicated that the quality of the established model is not only related to whether the characteristic wavelength is extracted, but also has a great relationship with the grouping method. It is of great significance to select the appropriate grouping method of training set and test set for the research object to improve the accuracy of the model. This provides a theoretical basis for future research.
Infrared Physics & Technology 125 (2022) 104240
Available online 6 June 2022
1350-4495/© 2022 Elsevier B.V. All rights reserved.
Hyperspectral detection of sugar content for sugar-sweetened apples based
on sample grouping and SPA feature selecting methods
Jie Chen
a
, Tiecheng Bai
a
, Nannan Zhang
a
, Lixia Zhu
b
, Xiao Zhang
a
,
*
a
College of Information Engineering, Tarim University, Alar, Xinjiang 843300, China
b
College of Food Science and Engineering, Tarim University, Alar, Xinjiang 843300, China
ARTICLE INFO
Keywords:
Sugar-sweetened apple
Sugar content
Training set
Test set
Grouping
Successive projection algorithms
Sample Grouping
ABSTRACT
This paper aims to select the spectral feature grouping method and establish regression detection models based
on feature grouping and successive projection algorithms (SPA) feature optimization. At rst, the hyperspectral
images and sugar content of 174 apple samples were collected to extract the hyperspectral data using ENVI 4.7
software. After pretreating the spectral data by using multiple scattering correction (MSC), all samples were
stored according to sugar content. In the case that the training set and the test set were divided into different
groups at the same allocation ratio, the full-band grouping modeling method and the continuous projection
extraction feature wavelength grouping modeling method were used to model the hyperspectral data and the
sugar content data, respectively. By comparing the prediction effect in terms of full-band group modeling in
which we character the grouping method as 1, 2, and 3, respectively. In the case of grouping method 1, i.e., when
the test set and training set were randomly assigned according to the set ratio column, the prediction effect of the
model established was optimal. Nonetheless, the simulation result was stable because it was the averaging R of
100 running results. In the case of grouping method 2, i.e., when the training set and the test set were divided
using interval allocation according to the set ratio, the prediction effect of the model, was slightly worse than that
of grouping method 1. Grouping method 3, i.e., the training set and the test set were grouped in series according
to the set ratio, has the worst prediction effect of the model. Results indicated that the quality of the established
model is not only related to whether the characteristic wavelength is extracted, but also has a great relationship
with the grouping method. It is of great signicance to select the appropriate grouping method of training set and
test set for the research object to improve the accuracy of the model. This provides a theoretical basis for future
research.
1. Introduction
Growing at the foot of the snow mountains, the Fuji apple trees of
Aksu are irrigating the melting-snow water all year long, and the diurnal
temperature in this region is large and there is almost no precipitation all
year round as well [16,23]. As a result, there is fructose condensed in-
side the apples which are commonly known as sugar-sweetened apples.
Additionally, the sugar-sweetened apple trees grow in the sandy soil on
the edge of the Taklamakan Desert, absorbing plenty of sunlight. The
large diurnal temperature makes the accumulation of organic matter in
photosynthesis of fruit trees much larger than respiration [3]. This ex-
plains why Aksu apples have much more sugar than those in other re-
gions. Aksu sugar-sweetened apple is known as the fruit queen of
Xinjiang, which is well received and welcomed by consumers due to its
good taste. Usually, it is common for Aksu apples to be fully booked-up
before their ripeness and picking each year, and the price is relatively
high as well. As a result, the lack of production could not meet the
consumers needs, inducing a large number of fake products in the
market. Although these fake apples look very similar in the prole, they
contain less sugar and taste worse conspicuously [12]. This can not only
affect the reputation and quality image of sugar-sweetened apples but
also cause a serious infringement of the rights as well as the damage to
interests of the consumers. Therefore, it is imperative to study sugar-
sweetened apples and their qualities.
Studies have shown that an apple doesnt crystallize until its sugar
content reaches the amount of 18%, implying that the sugar level is the
representative parameter of the sugar-sweetened apple, and the sugar
content is consequently taken as the quality parameter to study [10]. In
* Corresponding author.
E-mail address: zhangxiaoscnu@163.com (X. Zhang).
Contents lists available at ScienceDirect
Infrared Physics and Technology
journal homepage: www.elsevier.com/locate/infrared
https://doi.org/10.1016/j.infrared.2022.104240
Received 28 February 2022; Received in revised form 19 April 2022; Accepted 3 June 2022
Infrared Physics and Technology 125 (2022) 104240
2
recent years, spectral analysis has been widely used in agricultural
products and food quality as well as industrial analysis. An example of
this is near-infrared (NIR) spectroscopy, which could be used for non-
destructive testing of agricultural products and food, achieving good
results [1]. However, since the NIR spectrum is collected at a single
point, the collection data of apples and other samples with uneven
overall color distribution are incomprehensive enough [13,8]. There-
fore, the reliability of the established model is not good enough.
Hyperspectral technology is a new technology that combines image
and spectral techniques [17]. When irradiating an apple with the hyper
spectral, the light is reected from its surface in a diffuse reection,
hence the entire spectral image information of the object to be measured
can be obtained [2,14,9].
Hyper spectrum can also be used to detect the hardness, sugar con-
tent, moisture content, damage degree and other quality parameters of
samples. The mean spectrum of the region of interest has been studied
and a consequent sugar prediction model of apples has been established
by using the least square method [6]. The prediction model of apple
hardness and sugar content has been established using the Lorentz
parameter of hyper spectral images [4]. The hyper spectral imaging
technology and the principal component analysis has been used for
analyzing the wind injury and compression injury of apples, comparing
and analyzing the inuence of principal component analysis in different
spectral regions on the identication results [15]. The hyper spectral
apple images as well as the segmented regions of interest have been
collected and the spectral information has been extracted, additionally
the continuous projection algorithm has been used to extract charac-
teristic wavelengths and establish a model for apple disease detection
(Liu et al. 2017). These studies demonstrate the feasibility of combining
hyper spectral vision techniques with conventional spectral analysis
techniques for the detection of food and agricultural products. Actually,
hyper spectral technology develops rapidly due to its characteristic.
All of the above studies, however, have used xed training set and
test set distribution, without considering whether the distribution of the
two sets would affect the nal prediction effect. In this paper, the Aksu
Sugar-sweetened Apple in Southern Xinjiang was considered. At rst,
the hyperspectral images and sugar content of sugar-sweetened apples
were collected in a wavelength range of 900 nm to 1,800 nm. Then the
hyperspectral data was extracted using ENVI4.7, preprocess the spectral
data using multiple scattering correction, and sort the sugar content
(descending order or ascending order will not affect the nal result). The
method of full-band grouping modeling and the method of successive
projection algorithm (SPA) to extract feature wavelength groupings to
model the hyperspectral data and the confectionery data were used at
the same set ratio, as well as to make a comparative analysis of the nal
results.
2. Material and methods
In this paper we select the Aksu Sugar-sweetened Apples as the
experimental object, with all samples uniformed in size, shape, color
distribution, and having smooth surface without damage. We keep the
apples in the refrigerator after the purchase, then before the experiment,
we take out 180 ones, and place them at room temperature for 24 h and
numbered each of them in order. Having done this, we collect the hyper
spectral images and sugar content.
2.1. Acquisition of hyper spectral images
When taking hyper spectral images, we take three images of apples at
a time. A distance of 23 cm should be set apart between apples to avoid
not only interference of hyperspectral data of each apple but also in-
uence on later correction processing as well. We obtain the hyper-
spectral image data of sugar-sweetened apple samples by using
hyperspectral imaging system. The acquisition system is composed of
hyperspectral separator, enhanced near infrared hyperspectral camera
and standard white board. In order to avoid the environmental and
astigmatic interference during hyperspectral image acquisition, we
place the entire hyperspectral data acquisition system in a customized
black box. After releasing each sample, the gate can be closed, and then
the image can be collected by computer software. Fig. 1 shows the
hyperspectral image of an apple.
In the process of hyperspectral image acquisition, some noise infor-
mation will be included due to the difference of intensity distribution of
light source in various bands and the inuence of dark current noise of
camera alike. As this noise information can affect the quality of hyper-
spectral images, there are some errors in the acquisition of hyperspectral
data, hence the accuracy of the model will be greatly affected. So it is of
great importance that the hyperspectral images acquired must be cor-
rected for the purpose to reduce the impact of such interference on the
accuracy of the model. In this paper we use the software ENVI4.7. After
opening the le of the image of the sweet apple collected by the high
spectrometer, we use the white board for correction. Then we select the
part of the images to be processed in the equatorial region away from the
spot. Spectral data can be obtained by controlling the picture frame with
mouse buttons, and its spectral data are averaged and saved in the Excel
table in the form of ASCII. The spectral curve is shown in Fig. 2.
2.2. Collecting sugar content
We use a sugar salinity meter (MAST-BX /S28M) to measure the
sugar content of an apple at the location where spectral data are
collected illustrating in Fig. 1. We collet the sugar content 5 times at
each location and take the averaged value.
2.3. Data preprocessing
Hyperspectral data were collected from 200 samples, and 174 groups
of effective data were obtained by removing the data with large varia-
tion. The hyperspectral images of the original data are shown in Fig. 3.
After black and white correction, the hyperspectral image still contains
some noise. Therefore, some preprocessing methods will be needed to
further remove noise from hyperspectral images.
There are many ways to preprocess data. In this paper we analyze 11
pretreatment methods as illustrated in detail in Fig. 4. By comparison,
multiple scattering corrections (MSC) have the best preprocessing effect
Fig. 1. Image of sugared apple hyperspectral.
J. Chen et al.
Infrared Physics and Technology 125 (2022) 104240
3
on hyperspectral data of sugar-sweetened apple. MSC Pretreatment
method is a multi-variable scattering correction technique [11] which
has been used to preprocess data in many papers. The processed spectral
data can effectively eliminate the effect of scattering. Enhanced spectral
absorption information relates to component content, and images pre-
processed by MSC are shown in Fig. 5.
3. Results
3.1. Grouping of training sets and test sets
The grouping portion of the developed system is shown in Fig. 6. The
spectral data are stored in Excel which are processed in the system. The
quality parameters and spectral data are stored in an Excel table in
column format. It could be read by inputting the quality parameter of
the column number in the system. The preprocessing method could be
set in the parameter setting area (The selected number here corresponds
to the pretreatment method in Fig. 4). The grouping method of the
training set and test set could be set here, as well as the ratio of the
training set and test set. The system provides four grouping modes of
training sets and test sets. The number 1′′ means randomly assigned test
set and training set, and the assigned ratio could be set in the system.
The number 2′′ means that the training set and the test set are allocated
between each other, that is, according to the set ratio, to select one test
data from every number of training data. The 3′′ indicates the form
grouping in series the training set and the test set. According to the set
ratio, all the previous data are used as the training set, while the latter
data are used as the test set.
3.2. Data modeling and prediction
Common modeling methods include multiple linear regression,
principal component analysis, principal component regression, partial
least squares (PLS), articial neural network (ANN) and support vector
machine (SVM) [22]. PLS is an advanced multivariate statistical analysis
method. It is mainly used to solve the problem of multiple correlation
among variables in multiple regression analysis. Because it integrates
the basic skills of multivariate linear regression analysis, i.e., principal
component analysis and typical correlation analysis, it is known as the
second generation multivariate statistical analysis method, which is
Fig. 2. Image extraction spectrum collected by ENVI4.7.
Fig. 3. Original spectra extracted.
Fig. 4. Preprocessing method in the system.
Fig. 5. Curve of MSC after pretreatment.
J. Chen et al.
Infrared Physics and Technology 125 (2022) 104240
4
widely used in industrial design, metrology and econometrics [20,5,7].
The methods we use in this paper are the full-band grouping method and
the SPA grouping method. PLS is used in both of the above methods. The
full-band means that all the bands are adopted and no wavelength is
discarded, so the modeling speed and prediction speed will be slightly
slower. SPA can perform preliminary compression on hyperspectral
data, further screen the optimized wavelength according to its contri-
bution value to Sugar content, eliminate the insensitive wavelength,
reduce the complexity of the model, select the characteristic wave-
length, and predict the speed quickly [19,18,21].
All the pretreatment methods in this paper use method 3, namely
MSC pretreatment method. The ratio column of training set and test set
was set to 9, i.e., the ratio of training set and test set was 9:1. Fig. 6 shows
the prediction of the model when the full-band grouping modeling
method is grouped into method 1, i.e., when the test set and training set
are randomly assigned. Fig. 7 is the prediction of the model of interval
allocation between the training set and the test set according to the set
ratio when the full-band grouping modeling method is divided into
method 2. Fig. 8 shows the prediction of the model when using method 3
as the full-band grouping modeling method, i.e., when the training set
and the test set are grouped in series according to the set ratio. For the
full-band grouping modeling, according to the comparison of the above
Fig. 6. Prediction effect of grouping method 1.
Fig. 7. Prediction effect of grouping method 2.
J. Chen et al.
Infrared Physics and Technology 125 (2022) 104240
5
four models, it can be found that the prediction effect of the grouping
method 1 model is the best, with the correlation coefcient R =
0.916667. The prediction effect of the method 2 is slightly worse than
that of the method 1. Note that method 1 is the average value of 100
simulation results. In the case of method 3, i.e., training set and test set
are grouped in series according to the set ratio, the prediction effect of
the model is the worst. See Table 1 for the specic parameters.
Fig. 9 shows the prediction of the model which characteristic
wavelengths are extracted by SPA grouping method 1, and the number
of extracted characteristic wavelengths is 5, which are 1588, 1335, 923,
939 and 1732, respectively. Fig. 10 shows the prediction of the model
which characteristic wavelengths are extracted by SPA grouping as
method 2, and the number of extracted characteristic wavelengths is 5,
which are 1564, 1335, 923, 1743 and 1739, respectively. Fig. 11 shows
the prediction of the model which characteristic wavelengths is
extracted by SPA grouping method 3, and the number of extracted
characteristic wavelengths is 5, which are 1606, 1335, 923, 939 and
907, respectively. For SPA extraction feature wavelength grouping
modeling, when using method 2, the prediction effect of the model is the
best, with correlation coefcient R =0.87779. When using method 1,
the correlation is suboptimal, it is also the average of 100 simulations
results. And the effect of method 3 is the worst, with the correlation
coefcient of R is only 0.12798.
Combining the two modeling and prediction methods, the correla-
tion coefcient of the full-band grouping method will be closely to 1
when it is method 1, due to its randomness. We believe that method 2
will be more stable, and the correlation coefcient of the SPA extraction
feature wavelength grouping method will merely be suboptimal when it
is method 2. However, the former is not convincing enough as all data
are involved in modeling and prediction, so it is considered that the SPA
extraction feature wavelength grouping method will be optimal when it
is grouped using method 2. Fig. 12 is the comparison of measured value
and predicted value when the grouping modeling method of SPA
extracting characteristic wavelength is 2.
4. Conclusion
In this paper, after comparing 11 pretreatment methods, MSC
(multiple scattering correction) was used to preprocess the hyper-
spectral data of 174 groups of Aksu rock sugar core apples, and then 9:At
last, codes 1′′, 2′′ and 3′′ represent different grouping methods. Full-
band and SPA feature wavelength extraction methods are respectively
used to establish models and make predictions. The nal prediction
results show that: For the two modeling methods, when the whole band
group was 1′′, the correlation coefcient was 0.916667, and when the
characteristic wavelength extracted by SPA was 2′′ , the correlation
coefcient was only second. In other words, when the training set and
the test set were allocated according to the set proportion, the model
established by the feature wavelength extracted by SPA had a good
prediction effect. Therefore, the quality of the established model is not
only related to whether the characteristic wavelength is extracted, but
also has a great relationship with the grouping method. It is of great
signicance to select the appropriate grouping method of training set
and test set for the research object to improve the accuracy of the model.
This provides a theoretical basis for future research.
Fig. 8. Prediction effect of grouping method 3.
Table 1
Comparison of prediction effects.
Method Pretreatment method Ratio grouping method R Precision MSE
All band grouping MSC 9 1 0.916667 0.947521 0.826654
2 0.916570 0.947653 0.832416
3 0.739222 0.844863 2.648571
SPA characteristic wavele 1 0.863881 0.936293 1.05812
2 0.87779 0.93071 1.0107
3 0.12798 0.84081 2.8888
J. Chen et al.
Infrared Physics and Technology 125 (2022) 104240
6
Fig. 9. Effect diagram of SPA extraction feature wavelength grouping modeling method grouping as method 1.
Fig. 10. Effect diagram of SPA extraction feature wavelength grouping modeling method grouping as method 2.
Fig. 11. Effect diagram of SPA extraction feature wavelength grouping modeling method grouping as method 3.
J. Chen et al.
Infrared Physics and Technology 125 (2022) 104240
7
Declaration of Competing Interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Acknowledgement
Key Laboratory of Agricultural Engineering in Southern Xinjiang of
Tarim University (TDNG20180301); National Natural Science Founda-
tion of China (31960503); Key scientic and technological projects of
Xinjiang Construction Corps (2018AB042); Joint Funds of Tarim
university-China Agriculture University (ZNLH202102).
Availability of data and materials
All data are available from the corresponding author.
References
[1] S.X. Fan, Q.Y. Wang, Y.S. Yang, Development and experiment of fruit sugar content
vision-near infrared spectroscopy handheld device, Spectroscopy Spectral Anal. 41
(10) (2021) 6.
[2] T. Bai, T. Wang, Y. Chen, Comparison of Near-Infrared Spectrum Pretreatment
Methods for Jujube Leaf Moisture Content Detection in the Sand and Dust area of
Southern Xinjiang, Spectroscopy Spectral Anal. 39 (4) (2019) 1323.
[3] Jie Chen, Na Yao, Heng you Wang, Xiao Zhang, Hai fang Lv, Determination of
sugar content in apple by hyper spectrum with interval sampling algorithm for
sample subset and Ga-Pls model, Converter Mag. 21 (8) (2021) 310317.
[4] D. Feng, J. Ji, L. Zhang, Progress of hyperspectral imaging detection technology in
apple quality, Sci. Technol. Food Ind. 10 (2017) 389394, https://doi.org/
10.13386/j.issn1002-0306.2017.10.067.
[5] L.-N. Guo, C. She, D.-B. Kong, S.-L. Yan, Y.-P. Xu, M. Khayatnezhad, F. Gholinia,
Prediction of the effects of climate change on hydroelectric generation, electricity
demand, and emissions of greenhouse gases under climatic scenarios and
optimized ANN model, Energy Rep. 7 (2021) 54315445.
[6] Z. Guo, W. Huang, Y. Peng, X. Wang, J. Li, Impact of region of interest selection for
hyperspectral imaging and modeling of sugar content in apple, Mod. Food Sci.
Technol. 8 (2014) 5963.
[7] R. Hou, S. Li, M. Wu, G. Ren, W. Gao, M. Khayatnezhad, F. gholinia, Assessing of
impact climate parameters on the gap between hydropower supply and electricity
demand by RCPs scenarios and optimized ANN by the improved Pathnder (IPF)
algorithm, Energy. 237 (2021) 121621.
[8] D. Huang, J. Wang, M. Khayatnezhad, Estimation of actual evapotranspiration
using soil moisture balance and remote sensing, Iranian J. Sci. Technol., Trans. Civ.
Eng. 45 (4) (2021) 27792786.
[9] X. Li, X. Yue, Q. Huang, B. Zhang, Effects of wet-media milling on multi-scale
structures and in vitro digestion of tapioca starch and the structure-digestion
relationship, Carbohydr. Polym. 284 (2022), 119176, https://doi.org/10.1016/j.
carbpol.2022.119176.
[10] W.T. Liu, Nondestructive testing of apple quality based on hyperspectral imaging
technology [Diss], Agricultural University of Hebei, 2016.
[11] Y.J. Lu, Y.L. Qu, Z.Q. Feng, M. Song, Research on the method of choosing optimum
wavelengths combination by using multiple scattering correction technique,
Spectroscopy Spectral Anal. 27 (1) (2007) 5861.
[12] Q.L. Meng, J. Shang, X. Yang, Hyperspectral nondestructive testing of apple
hardness based on BP network, Packaging Eng. 41 (15) (2020) 59.
[13] Z.M. Qiao, C. Zhan, Prediction of apple sugar content based on NEAR infrared
spectroscopy and SSA-ELM, Food and Machinery 9 (2021) 121126.
[14] L. Su, W. Shi, X. Chen, L. Meng, L. Yuan, X. Chen, G. Huang, Simultaneously and
quantitatively analyze the heavy metals in Sargassum fusiforme by laser-induced
breakdown spectroscopy, Food Chem. 338 (2021), 127797, https://doi.org/
10.1016/j.foodchem.2020.127797.
[15] M. Sun, X. Chen, H. Zhang, H. Chen, Nondestructive inspect of apple quality with
hyperspectral imaging, Infrared Laser Eng. 4 (2014) 12721277.
[16] C. Wang, Y. Shang, M. Khayatnezhad, Fuzzy stress-based modeling for probabilistic
irrigation planning using Copula-NSPSO, Water Resource Manage. 35 (2021)
49434959, https://doi.org/10.1007/s11269-021-02981-6.
[17] H.Y. Wang, X.F. Li, Y.B. Li, Nondestructive testing of apple multi-quality
parameters based on hyperspectral image and 3D-CNN, J. Nanjing Agric. Univ. 431
(8) (2020).
[18] H. Wang, M. Khayatnezhad, N. Yousse, Using an optimized soil and water
assessment tool by deep belief networks to evaluate the impact of land use and
climate change on water resources, Concurrency and Computation 34 (10) (2022),
https://doi.org/10.1002/cpe.6807.
[19] S.L. Wang, L.G. Wu, N.B. Kang, H.Y. Li, J.Y. Wang, X.G. He, Study on tan-lamb
mutton tenderness by using the fusion of hyperspectral spectrum and image
information, J. Optoelectronics Laser. 27 (9) (2016) 987995.
[20] J. Wei, H. Zhao, Z. Gao, Multivariate statistical analysis in industrial design,
Modern Agric. Sci. 11 (2018) 816.
[21] J. Zhang, M. Khayatnezhad, N. Ghadimi, Optimal model evaluation of the proton-
exchange membrane fuel cells based on deep learning and modied African vulture
optimization algorithm, Energy Sources Part A 44 (1) (2022) 287305.
[22] J.J. Zhang, G.S. Liu, Y.C. Ren, W.H. Su, N.B. Kang, C. Ma, Rapid Detection of
Freshness in Tan-Lamb Mutton Based on Hyperspectral Imaging Technology,
Spectroscopy Spectral Anal. 39 (06) (2019) 19151921.
[23] K. Zhu, L. Liu, S. Li, B.o. Li, M. Khayatnezhad, A. Shakoor, Morphological method
and molecular marker determine genetic diversity and population structure in
Allochrusa, Caryologia 74 (2) (2021) 121130.
Fig.12. A comparison of measured and predicted values.
J. Chen et al.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Hydrological uncertainties are the main components of a sustainable framework in agricultural water management. Prediction of drought as a meteorological phenomenon should be considered to define the groundwater exploitation strategies. This study was conducted to develop a multiobjective-bivariate structure for reducing the soil moisture deficit and groundwater withdrawal in the Qazvin Irrigation District, Qazvin province, Iran. Therefore, non-dominated sorting theory, self-organizing particle swarm optimization and bivariate copula functions were incorporated under fuzzy uncertainty analysis. The results showed that the generalized extreme values and log-normal distribution functions had the best fitness on the drought peak and severity with Kolmogorov Smirnov amounts of 0.08 and 0.17, respectively. Furthermore, the goodness-of-fit tests were indicated the Joe joint function (MLE = 11) is the appropriate function for estimating the probabilistic values of drought characteristics. Proposed plans were to increase the water use efficiency for improving the expected yield production by an average of 20%. Furthermore, the standardized groundwater index was decreased from 1.1 to –4.3 for winter crops.
Article
Full-text available
In this study, an attempt is made to manage the gap between energy demand and energy supply by predicting hydropower production, energy demand, and greenhouse gas emissions. The interaction between climatic, hydrological, and socio-economic parameters creates a nonlinear and uncertain relationship. The complexity of this nonlinear relationship necessitate the use of ANN to estimate energy demand. To predict energy demand, ANN model is used along with improved Electromagnetic Field Optimization (IEFO) algorithms. The results show, hydroelectric generation in the near future under RCP2.6, RCP4.5, and RCP8.5 is decreased 10.981 MW, 12.933MW, and 14.765MW and in the far future decreased 21.922 MW, 23.649 MW, and 26.742 MW. The energy demand increases in the near future 513 MW and far future 1168 MW. According to forecasting hydropower generation and energy demand, the gap between the demand-supply will increase. Also, the greenhouse gases emissions is increase due to the increase in fossil fuel consumption.
Article
Full-text available
Determination of actual evapotranspiration is an important step for simulating the crop growth models in agricultural decision-making process. In this study, actual evapotranspiration (ETa) values for the main crops cultivated in Shahrekord plain, Chaharmahal and Bakhtiari province, Iran, are estimated based on real-time modeling and compared with the results obtained from the remote sensing data. Soil moisture balance was simulated considering the variation rate of crop canopy cover and its effects on evaporation and transpiration parameters. Daily values of ETa were predicted in a repeatable cycle based on the available soil moisture in root zone. Moreover, remote sensing images were analyzed as a data source to achieve the crop water requirement in critical times. For this objective, surface energy balance algorithm for land (SEBAL) was used to characterize the evapotranspiration and then derived ETa. Preliminary results obtained by applying satellite information showed a good agreement between the SEBAL estimates of evapotranspiration fluxes and the corresponding values simulated by the soil moisture estimation technique.
Article
Full-text available
As a nutritious and popular seafood among consumers, Sargassum fusiforme is susceptible to the toxic heavy metals because of its strong adsorption properties. In this study, laser-induced breakdown spectroscopy (LIBS) coupled with a simple framework (only remove some noise and low-intensity variables, and then combine with PLS algorithm) was used to establish the detection models to simultaneously and quantitatively analyze the content of heavy metals arsenic (As), chromium (Cd), cadmium (Cr), copper (Cu), mercury (Hg) , lead (Pb) and zinc (Zn) in Sargassum fusiforme. As comparisons, three classic variable methods of successive projections algorithm (SPA), uninformative variable elimination (UVE) and variable importance in projection (VIP) were adopted. The final results showed that six of seven heavy metal models from the TV-PLSR model were optimal. These results demonstrate that the TV-PLSR framework combined with LIBS technique is an effective framework for quantitatively analyzing the heavy metals in Sargassum fusiforme.
Article
A new optimized design of a hybrid AlexNet/Extreme Learning Machine (ELM) network to provide an optimal identification tool for the Proton-exchange membrane fuel cells (PEMFCs) is presented in this study. The major concept is to reduce the error amount between the empirical output voltage and the evaluated output voltage of the PEM fuel cell stack model using the proposed hybrid AlexNet/ELM. For enhancing the model formation of the AlexNet/ELM, a modified version of the African Vulture Optimization (MAVO) Algorithm, which is a new metaheuristic, is suggested. To analyze the efficiency of the suggested method, it is applied to a practical PEMFC benchmark case study for identification purposes. Then, the method is confirmed by comparison of the experimental data and standard AlexNet/ELM. The achievements indicated the better confirmation of the suggested AlexNet/ELM network with the experimental data. The results show that the highest relative error for training and test is 0.03% and 0.05342%, respectively, which shows a promising result for the study.
Article
This work aims to understand the relation between multi-scale structural changes and in vitro digestion of tapioca starch modified by wet-media milling. The native starch granules were smooth and difficult to digest due to the hindrance of enzymes entry into the granules by the concentric shell architecture. However, 1-min milling could break the smooth surface of starch granules, weakening shell architecture and increasing k1 (digestion rate). Moreover, 5–15-min milling could greatly disrupt the short-range ordered structure and transform it into the amorphous phase to enhance k1. The increase of milling time to 30–420 min mainly interrupted the glycosidic bonds of the amorphous phase, allowing the production of amylose and amylopectin fragments, achieving continuous increase of k1. Despite the increasingly severe damage caused by milling to multi-scale structures, the disruption of shell barrier on the starch surface was the most critical for digestion, followed by short-range ordered structure breakage.
Article
This article investigates the negative effect of land use and climate changes on water resources by the SWAT and SWAT-DEEP∖LMSFO model. Due to the importance of runoff impact on water resources in this article, the hybrid hydrological–deep neural networks optimized by the improved SFO based on logistic map (LMSFO) algorithm have been used to provide more accurate results for runoff estimation. This method improves runoff simulation. Firstly, runoff under the influence of land use and climate change is estimated by the SWAT model. Once again, runoff is estimated by the SWAT-DEEP/LMSFO model. In the DEEP/LMSFO model, the primary runoff has been estimated by an un-calibrated SWAT model. Then the Primary runoff simulated is entered as an input into the DEEP model. Finally, the runoff is estimated by a test-training method. The results of the SWAT model and the DEEP/LMSFO model show that there is an inverse correlation between land use so that reducing land cover increases runoff. The results show that climate change and land-use change can affect annual runoff changes in the coming years.
Article
The Caryophyllaceae family is complex. Several attempts have been carried out in the past to study Caryophyllaceae members. This study mainly focused on Allochrusa Bunge to determine its genetic structure and used ISSR markers, ITS, and rps16 data to classify and differentiate Allochrusa species. We collected 122 Allochrusa specimens. Our analysis included morphological and molecular method approaches. Morphometry analysis indicated that floral characters could assist in the identification of Allochrusa species. A. persica (Boiss.) Boiss. and A. versicolor Fisch. & C.A.Mey. showed affinity to each other. A. bungei Boiss. formed a separate group. Analysis of molecular variance showed significant genetic differentiation in Allochrusa (p= 0.001). The majority of genetic variation was among the Allochrusa population. We recorded minimum gene flow (Nm=0.176) between Allochrusa species. Besides this, isolation by distance occurs in Allochrusa members, as shown in the Mantel test result (r = 0.01, p = 0.0002). STRUCTURE analysis revealed three genetic groups. It is evident that A. persica, A. versicolor, and A. bungei differ genetically from each other. Our current findings have implications in plant systematics and biodiversity management.
Article
This study evaluates the effect of climate change on electricity generation, electricity demand, and GHG emissions. For this purpose, using climate scenarios RCPs changes of climatic parameters are predicted. Due to the high importance of energy demand in the management of energy generation resources innovation research is related to forecasting electricity demand. The novelty is the use of an Artificial Neural Network optimized to predict the energy demand. To optimize the ANN method, the Improved Pathfinder algorithm has been used. The use of the optimization method in the ANN method provides a model with more precision and fewer errors for the prediction of energy demand. The results showed that due to the weather changes, hydropower generation for the near future under RCP2.6, RCP4.5, and RCP8.5 increases by about 2.765 MW, 1.892 MW, and 1.219 MW and for the far future increases by about 3.430 MW, 2.475 MW, and 1.827 MW. The electricity demand forecasting by The ANN-IPF model for the near and far future will increase compared to the base period of 391.9 MW and 716.65 MW, respectively. Therefore, the gap between the demand the power supply will increase. Using other resources, the difference between demand and power supply will decrease.
Article
The freshness of Tan mutton is an important index of its quality and safety, and it is also a key link in the quality control of meat products. Total Volatile Basic Nitrogen (TVB-N) is the main chemical information which can effectively reflect the loss of freshness of Tan mutton. However, the traditional detection method of TVB-N must destroy the samples, the detection process is tedious, the man-made influencing factors are large, and the test result is lack of objectivity and consistency. Hyperspectral imaging technology which is a non-destructive method meets the needs of modern detection technologies for multi-source information fusion that has been widely used in the field of food safety. This paper used visible/near-infrared spectroscopic imaging technology (400~1 000 nm) combined with dynamics and chemometrics methods and computer programming to achieve the rapid detection of TVB-N concentration and prediction of safe storage period during the storage period of Tan mutton (15 ℃). The research contents were as follows: The average spectral data for each sample area of interest were extracted and the monte carlo algorithm was selected to eliminate the abnormal samples. The X-Y symbiotic distance (Sample set partitioning based on joint X-Y distances, SPXY) was used to divide the mutton set into the correction set and the prediction set. Multiplicative Scatter Correction (MSC), Savitzky-Golay (SG), Standard Normalized Variate(SNV), normalization (Normalization) and baseline calibration (Baseline) were used to preprocess the original spectral data. 21 and 6 feature wavelengths were extracted by the Campetitive Adaptive Reweighted Sampling (CARS) and Successive Projections Algorithm (SPA). In order to simplify the model and improve the accuracy of prediction of the model, the SPA algorithm was used to perform secondary extraction of selected feature wavelengths of CARS and 14 feature wavelengths were selected. A PLSR model with TVB-N concentration was established based on the extracted characteristic wavelengths, and the SNV-CARS-SPA-PLSR model was preferred to have a higher prediction ability (R2c=0.88, RMSEC=2.51, R2p=0.65, RMSEP=2.11) Meanwhile, a dynamic model of mutton TVB-N change and storage time could be established. Finally, the dynamic model of spectral absorbance value and storage time of mutton were established by combining the optimized spectral model with the dynamic first order reaction model, and predicte the storage time, and the PLS-DA model was realized to discriminate the storage time of mutton (the correction set discriminant accuracy rate was 100%, and the prediction set is 97%). The result showed that visible/near-infrared hyperspectral imaging technology in combination with dynamics and chemometrics methods and computer programming could effectively detect TVB-N index of mutton rapidly and non-destructively, and be realized to monitor the quality and safety of mutton and provide a theoretical reference for developing on line defection equipment.