research articledownloads.hindawi.com/journals/jfq/2018/4706147.pdfresearch article assessment of...

Research ArticleAssessment of Ripening Degree of Avocado by ElectricalImpedance Spectroscopy and Support Vector Machine

Monzurul Islam Khan Wahid and Anh Dinh

Department of Electrical and Computer Engineering University of Saskatchewan Saskatoon S7N 5A2 Canada

Correspondence should be addressed to Monzurul Islam moi352mailusaskca

Received 7 May 2018 Accepted 2 September 2018 Published 1 November 2018

Academic Editor Antonio Piga

Copyright copy 2018 Monzurul Islam et al is is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited

Avocado a climacteric fruit exerts high rate of respiration and ethylene production and thereby subject to ripening duringstorage erefore its ripening is a signicant factor to impart optimum quality in postharvest storage To understand thedynamics of ripening and to assess the degree of ripening in the avocado electrical sensing technique is utilized in this study Inparticular electrical impedance spectroscopy (EIS) is found to uncover the physiological and structural characteristics in plantsand vegetables and to follow physiological progressions due to environmental impacts In this work we present an approach thatwill integrate EIS and machine learning technique that allows us to monitor the ripening degree of the avocado It is evident fromour study that the impedance absolute magnitude of the avocado gradually decreases as the ripening stages (rm breaking ripeand overripe) proceed at a particular frequency In addition principal component analysis shows that impedance magnitude (twoprincipal components combined explain 9995 variation) has better discrimination capabilities for ripening degrees compared toimpedance phase angle impedance real part and impedance imaginary part Our classier utilizes two principal componentfeatures over 100 EIS responses and demonstrates classication over rm breaking ripe and overripe stages with an accuracy of90 precision of 93 recall of 90 f1-score of 90 and auc of 88 e study oers plant scientists a low cost and non-destructive approach to monitor postharvest ripening process for quality control during storage

1 Introduction

Avocados receive an increasing attention for extendingnutritional food choice and agribusiness in United States [1]According to a recent report from the National AgriculturalStatistics Service of United States Department of Agricul-ture the value of US avocado production measured $316million in 2016-17 [2] and US consumption of avocadosincreased signicantly from 11 pounds per capita in 1989 toa record 71 pounds per capita in 2016 Avocado beinga climacteric fruit has a high rate of postharvest respirationConsequently it is one of the most perishable fruits and hasvery limited shelf life It is prone to biochemical andphysiological deterioration during postharvest ripeningaccompanied by degradation of visual appearance ispostharvest loss poses the risk of loss of market value of theavocado In addition from the point of view of consumerindustry only optimum ripening state attributes to the most

nutritional value and best taste of a fruit Hence a betterunderstanding of ripening dynamics of avocado can playa vital role to the development of appropriate tool for betterpackaging storage and transportation process and conse-quently meet the demand of both agribusiness and con-sumer industry

Conventional chemical and biochemical analyses con-ducted to investigate the fruit ripening are limited by factorssuch as processing time and destructive nature [3] esemethods are often laborious and expensive and require accessto laboratory facility Hence these methods prove to be in-feasible for a repetitive inspection erefore it is essential toexpand current technologies from dierent viewpoints Anondestructive low cost and easily accessible solution to thisissue needs to be devised

Nondestructive methods such as magnetic resonanceimaging (MRI) and CTscan are found to be eective towardsunderstanding of ripening internal fruit quality and

HindawiJournal of Food QualityVolume 2018 Article ID 4706147 9 pageshttpsdoiorg10115520184706147

postharvest processing [4 5] MRI method requires sepa-ration of signal of water proton from that of fat proton whichis still a challenging task And both MRI and CT are limitedby the factors such as high cost and processing complexityHyperspectral imaging can effectively assess ripening de-grees [6] but it also suffers from constrains such as cost andprocessing erefore it is essential to expand currenttechnologies from different viewpoints On the contraryelectrical impedance spectroscopy (EIS) is a fast low costand nondestructive method which is found to offer insightinto plant physiology and physiological dynamics due toenvironmental impacts In this direction the EIS studieshave been conducted as a nondestructive evaluation methodto investigate the impedance spectrum variations and todetermine the ripening degree in the avocado

To develop an easily accessible and nondestructivemethod to understand mechanisms of ripening and to assessripening degree of the avocado the prospect of EIS tech-nique is explored in this paper is work on the avocadomainly focuses on the investigation of feasibility of EIS forassessment of the ripening degree nondestructively e restof the paper is organized as follows related past works onapplication of EIS technique is reviewed in the LiteratureReview section e next section introduces the theory ofbioimpedance and electrical impedance spectroscopy Lateron experimental methods and materials are illustratedAfter that the experimental and simulated results are pre-sented and discussed and finally paper is concluded in-cluding some future exploration directions

2 Literature Review

Impedance sensing technology especially EIS has emergeda new era into food quality and stability over the last decadesIt has been extensively used in the field of plant physiologyagriculture and food engineering for quality control andassessment of fruits and vegetables such as banana [7] Garutcitrus [8] kiwi [9] lettuce [10] nectarine [11] and straw-berry [12] In order to assess the freshness of banana EISinvestigation was performed during different ripening states[7] By attaching AgAgCl electrode and injecting a smallamount of current impedance responses are measured bythe 4294A impedance analyzer over a frequency range of50Hz to 1MHz e impedance magnitude phase anglereal part and imaginary part varied markedly with the al-teration of the ripening state

Gonzalez-Araiza et al [13] designed a nondestructivedevice to obtain the impedance spectrum of the wholestrawberry fruit and later on performed classification byutilizing corresponding equivalent circuit parameters (con-stant phase element CPE-P and Rinfinity) e study showedthat the strawberries at the highest stage of ripeness hadsignificantly lower constant phase element and Ro (related toextracellular) values compared to other strawberries

Neto et al [14] utilized EIS technique for the de-termination of the maturation degree of mangoes based onvariation of bulk resistance dependence with maturation offruits ey came up with this strategy to normalize bulkresistance by diameter to compensate the size variations of

samples ey demonstrated good agreement between vari-ation of electrical response and mechanical response of fruit

Montoya et al [15] investigated that electrical conduc-tivity could be a suitable factor for assessment of qualityduring ripening and cold storage ey found some re-semblances between electrical conductivity response andethylene productions curve ey defined a threshold ofconductivity (024 Sm) that indicates the limiting value forfruit stored at noninjurious temperatures and subsequentlytransferred to 20degC for marketing

Chowdhury et al [16] carried out EIS study on themandarin orange fruit during ripening in a spectrum be-tween 50Hz and 1MHz ey observed significant variationof the impedance phase angle real part and imaginary partof the impedance with different states of orange ripeninge study also demonstrated loss of weight of correspondingsamples with the progression of ripening states uselectrical sensing specially EIS technique was found to offerinsight into physiology of fruits and vegetables undergoingthe ripening process

3 Theory of Bioimpedance and ElectricalImpedance Spectroscopy

e plant body is a complex biological structure composedof tissues which are developed with cells suspended in ex-tracellular fluids (ECF) [17] Again cells are composed ofintracellular fluids (ICF) cell membrane (CM) and cell wall(CW) ECF ICF and CM are developed with differentmaterials and so exhibit distinguishable electrical attributese ECF and ICF act as electrolytes and provide a con-ducting path to applied alternating current [18] e CM isa protein-lipid-protein (P-L-P) structure and exhibits ca-pacitance to the current Consequently the overall responseof the biological tissues to an alternating electrical signalgenerates a complex sbioelectrical impedance Mathemati-cally the impedance Zang(θ) is calculated by dividing thevoltage (Vang(θ1)) measured by applied current (Iang(θ2)) as

Zang(θ) Vang θ1( 1113857

Iang θ2( 1113857 (1)

Bioimpedance is a complex quantity which varies withtissue composition and frequency of the applied signalerefore the frequency-dependent bioimpedance can berepresented as

Zb(ω) Re(Z(ω)) minus jIm(Z(ω)) Rb(ω)minus jXb(ω) (2)

where Rb(ω) and Xb(ω) represent the magnitude of the realpart of the complex Zb(ω) and the magnitude of theimaginary part of the complex Zb(ω) respectively

Bioimpedance is sensitive to the physiological status ofplant tissue Again as the response of bioimpedancechanges with frequency a multifrequency impedanceanalysis can offer better insight into plant physiology andbetter understanding to plant tissue status Electricalimpedance spectroscopy (EIS) is a multifrequency analy-sis for studying complex electrical impedance Z(ω)and its phase angle θ(ω) at different frequency points

2 Journal of Food Quality

ωi(ωi ω1ω2ω3 ωn) EIS is performed by measuringthe surface potentials V(ω) occurring from a constantcurrent injection I(ω) at the boundary through a linear arrayof the surface electrodes attached to the sample-under test(SUT)

4 Materials and Methods

41 EIS Measurement Impedance measurement on theavocado was carried out with the EVAL-AD5933 EvaluationBoard a high precision impedance converter system edevice integrates an on-board frequency generator a 12-bit 1MSPS analog-to-digital converter (ADC) and an internaltemperature sensor Both the excitation signal and responsesignal are sampled by ADC and Fourier transformed by anon-board DSP engine in order to obtain complex impedancespectrum e frequency range of AD5933 is from 5 kHz upto 100 kHz without external components and lower fre-quencies than 5 kHz are achievable using an external dividere device has a master clock of 1677MHz and supplyvoltage requirement of 27 V to 55V e device comes ina 16-SSOP package that has a temperature range of minus40degC to+125degC e device offers high accuracy and versatility thatmake it suitable for electrochemical analysis corrosionmonitoring automotive sensors proximity sensing andbioimpedance measurements

e experimental setup of EIS data acquisition system ispresented in Figure 1(a) e graphical user interface of thesupporting software is presented in Figure 1(b) e electrodeholder platform has a height of 10 centimeter and the holderis horizontally movable and vertically rotatable A non-invasive two-electrode measurement system was employed inour experiment where ldquoalligator-rdquo type Ag electrode clipswere connected with circular electrode plates from ECGelectrode and finally ldquoalligatorrdquo clips were tied with theelectrode holders Similar separations of 44 centimeter be-tween two electrodes were maintained for all measurementsTo compensate the contact resistance a layer of electrode gel(ldquoSpectra 360rdquo salt-free electrode gel) was placed betweenECG electrode plate and test subject (avocado fruits) Forimpedance spectroscopy measurements in this study we useda 1V p-p generator voltage and scanned 101 spot frequencies(frequency intervals) between 5 kHz and 15 kHz e ACsignal injected into the sample was generated by the built-infunction generator of the evaluation board

e avocado fruits for our experiment were collectedfrom a local supermarket ldquoSobeysrdquo in Saskatoon Sas-katchewan Canada During the period of experiment av-ocados were kept in the lab environmente temperature inthe lab was 20degC and relative humidity was 40 Furtherexperiment can be conducted with freshly harvested avo-cados and stored in different controlled temperatures infuture studies

42 Feature Extraction by Principal Component Analysis(PCA) PCA is mathematically defined as an orthogonallinear transformation that converts a set of data (possiblycorrelated) into a new set of linearly uncorrelated data called

principal components e first principal componentcontains the largest percentage of data variance and thevariance decreases in the following principal componentsIn this study PCAs were carried out with data obtainedfrom the samples in order to assess the feasibility of the EIStechnique to discriminate among different ripening stagesof the avocado PCA served the purpose of dimensionalityreduction and feature extraction on EIS data PCA wasperformed over impedance magnitude impedance phaseangle impedance real part and impedance imaginary partover the predefined frequency range where the deviceshows maximum sensitivity

43 Classification by Multiclass Support Vector Machine(SVM) Support vector machines (SVMs) are supervisedlearning models which construct an optimal hyperplane toclassify data into different classes And lines drawn parallelto this separating line are the supporting hyperplanes andthe distance between them is called the decision marginewidth of the margin is constrained by support vectors whichare the data points that are closest to the separating hy-perplane Since the optimal hyperplane is the one thatseparates the high probability density areas of the classeswith maximum possible margin between them the goal is todetermine the direction that provides the maximummarginIt needs the solution of following optimization problem inEquation (3) for a given training set of instance label pairs(xl yl) l 1 2 i where xl isin Rn and yl isin 1minus1 i

minueξ

12u

Tu + C 1113944

i

l1ξl

subjected to yl uTϕ xl( 1113857 + e( 1113857ge 1minus ξl ξl ge 0

(3)

For quantitative analysis of discrimination of ripeningdegrees based on EIS data a multiclass support vectormachine was modeled e ground truths for all the sampleswere generated by subjective testing In this directionripening chart and ripening information of the avocado wereutilized e ripening chart of the avocado used in thisexperiment is presented in Figure 2

5 Results and Discussion

Ripening is the process by which fruits attain their desirablecolor flavor palatable nature and other textural propertiesthat make the fruit acceptable for consumption Ripening isassociated with change in biochemical composition that isconversion of starch to sugar e avocado being a cli-macteric fruit continues to ripen after harvest During theripening process it emits ethylene along with the increasedrate of respiration

e EIS study on the avocado was undertaken with anattempt to better understand and identify the ripeningstages of the avocado in terms of electrical bioimpedancee EIS studies conducted for the avocado sample showthat at a particular frequency (especially for less than10 KHz) the avocado impedance magnitude gradually

Journal of Food Quality 3

(a)

(b)

Figure 1 Hardware and software utilized for data acquisition (a) AD5933 evaluation board and (b) snapshot of the supporting softwarersquosgraphic user interface

Firm

Days to ripe 4ndash5 Days to ripe 1ndash2 Days to ripe 0 Days to ripe past dueNot ripe Almost ripe Ready to eat Past ripe

Breaking Ripe Overripe

Figure 2 Ripening chart of avocado


decreases as ripening stages (rm breaking ripe andoverripe) proceed (Figure 3(a)) And for a particularmaturity degree impedance magnitude decreased signi-cantly from low to high frequency especially for rmbreaking and ripe stages At lower frequency the electricalcurrent centows only through extracellular centuids which acts aselectrolytes and have relatively high resistance e cellmembrane exerts extreme high capacitance at low fre-quency and that is why electrical current cannot passthrough intracellular centuid and only centows through theextracellular centuid At high frequency cell membrane ca-pacitance reduces signicantly and current centows throughintracellular centuid which has relatively low resistance is

is how impedance declines markedly from low to highfrequency area of impedance spectra is phenomenonresulting from cell structures in biological tissue is knownas β dispersion Figure 3(b) shows that the phase angle ofavocado impedance varies with frequency throughout theripening stages At a particular frequency the avocado im-pedance phase angle gradually increases as ripening stages(rm breaking ripe and overripe) proceed (Figure 3(b))And especially on high-frequency range of the spectrum(10KHz to 15KHz) for a particular maturity degree theimpedance phase angle rises as the frequency increases

EIS study on the avocado was extended to multiplesamples to uncover the dynamics of ripening with respect to

05 06 07 08 09 1 11 12 13 14 15Frequency in Hz times104

05

1

15

2

25

3

Impe

danc

e mag

nitu

de in

ohm

times104

FirmBreaking

RipeOverripe

(a)


ndash70

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

Neg

ativ

e pha

se an

gle i

n de

gree

times104

FirmBreaking

RipeOverripe

(b)

Figure 3 EIS response of avocado during ripening (a) impedance magnitude vs frequency and (b) negative phase angle vs frequency


impedance-frequency response One hundred dierent EISresponses belonging to dierent ripening degrees wererecorded and are shown in Figure 4 It can be intuitivelyconcluded that impedance magnitude best recentects the rip-ening degrees

Now to better visualize and analyze the large data ob-tained from all EIS responses principal component analysis(PCA) was carried out over impedance magnitude im-pedance phase angle impedance real part and impedanceimaginary part Furthermore PCA was used to obtaina reduced number of uncorrelated variables which are theprincipal components (PCs) PCA was able to reduce the

initial 101 variables (for every impedance parameter) to 2PCs which combined and explained a signicant percentageof the total variance e rst principal component andsecond principal component over impedance magnitudeexplained respectively 9867 and 128 of total variancein the data PCA over impedance magnitude impedancephase angle impedance real part and impedance imaginarypart is demonstrated respectively in Figures 5(a)ndash5(d)Especially in the PCA2-PCA1 score plot of impedancemagnitude (Figure 5(a)) it is apparent that the data for everyripening degrees tend to cluster and in general four dif-ferent zones can be separated It is evident that PCA feature


times104

05

1

15

2

25

3

35

Impe

danc

e mag

nitu

de in

ohm

FirmBreaking

RipeOverripe

(a)


FirmBreaking

RipeOverripe

0

02

04

06

08

1

12

14

Neg

ativ

e pha

se an

gle i

n ra

dian

(b)

times104


FirmBreaking

RipeOverripe

04

06

08

1

12

14

16

18

2

22

Real

par

t of i

mpe

danc

e in

ohm

(c)

times104


FirmBreaking

RipeOverripe

0

05

1

15

2

25

3

35

Imag

inar

y pa

rt o

f im

peda

nce i

n oh

m

(d)

Figure 4 EIS response of all experimental avocado samples during ripening (a) impedance magnitude vs frequency (b) negative phaseangle vs frequency (c) real part of impedance vs frequency and (d) imaginary part of impedance vs frequency


from impedance magnitude has better discrimination ca-pabilities for ripening degrees compared to impedance phaseangle impedance real part and impedance imaginary part

In addition for quantitative investigation of the prospectof EIS parameters towards discrimination of ripening de-grees a supervised classier model was developed Ourexperimental dataset is composed of 100 EIS responsescontaining 19 responses from rm class 26 from breakingclass 26 from ripe class and 29 from overripe class Duringexperiment the database was divided into two sets thetraining set containing 60 responses and the testing setcontaining 40 responses For feature extraction from EISdata PCA was carried out over impedance magnitudeimpedance phase angle impedance real part and impedanceimaginary part By analyzing discrimination capabilities ofdierent impedance parameters only PCA1 and PCA2 overimpedance magnitude were selected to feed to our classierFor classication purpose a multiclass support vector ma-chine with ldquolinearrdquo Kernel was utilized For performanceevaluation of the classication model performance pa-rameters such as accuracy precision recall and F1-scorewere calculated Accuracy most intuitive performancemeasure is simply a ratio of correctly predicted observation

to the total observations Precision is the ratio of correctlypredicted positive observations to the total predicted posi-tive observations e recall is intuitively the ability of theclassier to nd all the positive samples e F-beta score canbe interpreted as a weighted harmonic mean of the precisionand recall At the 60ndash40 train-test split the testing ac-curacy of the classication was 90 e other performancemeasures are illustrated in Table 1

To assess the accuracy of a classication it is commonpractice to create a confusion matrix where classicationresults are compared to additional ground truth in-formation Figure 6 shows the graphical representation ofthe confusion matrix containing test data for rm (class 0)breaking (class 1) ripe (class 2) and overripe (class 3)

ndash4 ndash2 8 10PCA1

ndash15

ndash1

ndash05

0

05

1

PCA

2

times104

times104

PCA over Z magnitude

FirmBreaking

RipeOverripe

0 2 4 6

(a)

times104

times104

ndash08

ndash06

ndash04

ndash02

0

02

04

06

FirmBreaking

RipeOverripe

PCA

2

PCA over negative phase angle

ndash4ndash6 ndash2 8PCA1

0 2 4 6

(b)

times104

times104

ndash15

ndash1

ndash05

0

05

1

15

FirmBreaking

RipeOverripe

PCA1

PCA

2

PCA over real part of impedance

ndash3 ndash2 ndash1 4 50 1 2 3

(c)

times104

times104

ndash08ndash1

ndash06ndash04ndash02

002040608

1

FirmBreaking

RipeOverripe

PCA1

PCA

2

PCA over imaginary part of impedance

ndash4ndash6ndash8ndash12 ndash10 ndash2 80 2 4 6

(d)

Figure 5 Principal component analysis (PCA2 vs PCA1) over EIS responses (a) impedance magnitude (b) negative phase angle (c) realpart of impedance and (d) imaginary part of impedance

Table 1 Performance scores of proposed algorithm

ClassPerformance scores

Precision Recall F1-score0 rm 100 100 1001 breaking 100 080 0892 ripe 071 100 0833 overripe 100 083 091Averagetotal 093 090 090


A receiver operating characteristic (ROC) curve is cre-ated by plotting the true positive rate (TPR) against the falsepositive rate (FPR) at various threshold settings It illustratesthat the diagnostic ability of a classier system and the areaunder the ROC curve called AUC is an eective summa-rization of its performance Figure 7 shows that the averagearea under the ROC curve is 88 that indicates excellentdiscrimination capabilities of our classier

In 2011 a study was performed by Rehman et al [19] toinvestigate the ripening process of the mango utilizingelectrical impedance spectroscopy technique To discrimi-nate raw and ripe mango fruits they measured bulk im-pedance by LCR meter and expressed in terms of eectiveresistance and eective capacitance in the frequency range of

1 to 200KHz ey came up with a ratio of eective re-sistance of ripe to raw fruit that is signicant enough at1 KHz to characterize raw and ripe state ey also found theratio of eective capacitance of raw and ripe mango andconcluded that the ratio of eective resistance shows betterdiscrimination capabilities at 1ndash6KHz compared to eectivecapacitance In our study we dierentiated among fourripening states of the avocado utilizing impedance spectra at5 KHz to 15KHz utilizing AD5933 evaluation board isimpedance converter board is low cost compared to theLCR-800 GW Instek in their study Also it supports au-tomated frequency sweep which their LCR meter lacks efeature extraction by principal component analysis overimpedance magnitude directly (instead of eective re-sistance or capacitance) was enough for our support vectormachine classier As their study only dierentiated binaryclasses (raw and ripe) we performedmulticlass classicationamong rm breaking ripe and overripe avocado e studyoers plant scientists a low cost and nondestructive ap-proach to monitor postharvest ripening process of the av-ocado for quality control during storage

6 Conclusion

e feasibility of electrical impedance spectroscopy a non-destructive technique to assess the ripening degree of theavocado has been explored in this study A low cost easilyaccessible and nondestructive system based on AD5933impedance analyzer has been investigated in this work eelectrical impedance parameter especially impedance absolutemagnitude is found to be most sensitive to ripening pro-gression on the avocado In addition principal componentanalysis over frequency-dependent electrical response cor-roborates the hypothesis of distinguishing ripening statesbased on EIS Our classier based on multiclass supportvector machines shows excellent discriminant capabilities ofEIS technique to track and analyze 4 ripening stages (rmbreaking ripe and overripe) of the avocado is approachcan be a potential alternative to conventional chemicalanalysis techniques with oering of better time and costsaving and less processing complexity e proposed systemcan be extended and tested to other fruits for analyzing theirripening dynamics nondestructively and to be addressed infuture studies

Data Availability

e data used to support the ndings of this study areavailable from the corresponding author upon request

Conflicts of Interest

e authors declare that there are no concenticts of interestregarding the publication of this paper

References

[1] M Wien E Haddad K Oda and J Sabate ldquoA randomized3x3 crossover study to evaluate the eect of Hass avocadointake on post-ingestive satiety glucose and insulin levels and

Confusion matrix without normalization

Class 0 firm

Clas

s 0 f

irmClass 1 breaking

Clas

s 1 b

reak

ing

Class 2 ripe

Clas

s 2 r

ipe

Class 3 overripe

Clas

s 3 o

verr

ipe

Predicted label

8

8

8

6

4

2

2

2

10

10

10

0 0 0

0

0

0

0

0

0

0

0

True

labe

l

Figure 6 Plot of confusion matrix for test data

Some extension of receiver operating characteristic to multiclass10

10

08

08

06

06

04

04False positive rate

Microaverage ROCcurve (area = 088)ROC curve of class 0(area = 100)ROC curve of class 1(area = 085)

ROC curve of class 2(area = 067)ROC curve of class 3(area = 098)

True

pos

itive

rate

02

020000

Figure 7 ROC analysis of ripening stage classication


subsequent energy intake in overweight adultsrdquo NutritionJournal vol 12 no 1 p 155 2013

[2] NASS Noncitrus Fruits and Nuts 2016 Summary NationalAgricultural Statistics Service United States Department ofAgriculture Washington DC USA 2017

[3] K Miloski K Wallace A Fenger E Schneider andK Bendinskas ldquoComparison of biochemical and chemicaldigestion and detection methods for carbohydratesrdquo Amer-ican Journal of Undergraduate Research vol 7 pp 48ndash522008

[4] M K Abera S W Fanta P Verboven Q T Ho J Carmelietand BM Nicolai ldquoVirtual fruit tissue generation based on cellgrowth modellingrdquo Food and Bioprocess Technology vol 6no 4 pp 859ndash869 2013

[5] L Zhang and M J McCarthy ldquoMeasurement and evalua-tion of tomato maturity using magnetic resonance imag-ingrdquo Postharvest Biology and Technology vol 67 pp 37ndash432012

[6] R Khodabakhshian and B Emadi ldquoApplication of VisSNIRhyperspectral imaging in ripeness classification of pearrdquo In-ternational Journal of Food Properties vol 20 no 3pp S3149ndashS3163 2017

[7] A Chowdhury T Bera D Ghoshal and B ChakrabortyldquoStudying the electrical impedance variations in bananaripening using electrical impedance spectroscopy (EIS)rdquo inProceedings of 2015 ird International Conference on Com-puter Communication Control and Information Technology(C3IT) pp 1ndash4 Berhampore West Bengal India 2015

[8] J Juansah I W Budiastra K Dahlan and K B Seminar ldquoeprospect of electrical impedance spectroscopy as non-destructive evaluation of citrus fruits acidityrdquo InternationalJournal of Emerging Technology and Advanced Engineeringvol 2 pp 58ndash64 2012

[9] Y Tang G Du and J Zhang ldquoChange of electric parametersand physiological parameters of kiwi of storage periodrdquoNongye Jixie Xuebao(Transactions of the Chinese Society ofAgricultural Machinery) vol 43 pp 127ndash133 2012

[10] R F Muntildeoz-Huerta A d J Ortiz-Melendez R G Guevara-Gonzalez et al ldquoAn analysis of electrical impedance mea-surements applied for plant N status estimation in lettuce(Lactuca sativa)rdquo Sensors vol 14 no 7 pp 11492ndash115032014

[11] M D OrsquoToole L A Marsh J L Davidson Y M TanD W Armitage and A J Peyton ldquoNon-contact multi-frequency magnetic induction spectroscopy system forindustrial-scale bio-impedance measurementrdquo Measure-ment Science and Technology vol 26 no 3 article 0351022015

[12] J R G Araiza Impedancia Bio-Electrica Como Tecnica No-Destructiva para Medir la Firmeza de la Fresa (Fragaria xAnanassa Duch) y su Relacion Con Tecnicas ConvencionalesUPV Universitat Politecnica de Valencia Valencia Spain2014

[13] J R Gonzalez-Araiza M C Ortiz-Sanchez F M Vargas-Luna and J M Cabrera-Sixto ldquoApplication of electrical bio-impedance for the evaluation of strawberry ripenessrdquo In-ternational Journal of Food Properties vol 20 no 5pp 1044ndash1050 2017

[14] A F Neto N C Olivier E R Cordeiro and H P de OliveiraldquoDetermination of mango ripening degree by electrical im-pedance spectroscopyrdquo Computers and Electronics in Agri-culture vol 143 pp 222ndash226 2017

[15] M Montoya J D L Plaza and V Lopez-Rodriguez ldquoElec-trical conductivity of avocado fruits during cold storage and

ripeningrdquo LWT-Food Science and Technology vol 27 no 1pp 34ndash38 1994

[16] A Chowdhury P Singh T K Bera D Ghoshal andB Chakraborty ldquoElectrical impedance spectroscopic study ofmandarin orange during ripeningrdquo Journal of Food Mea-surement and Characterization vol 11 no 4 pp 1654ndash16642017

[17] T K Bera ldquoBioelectrical impedance methods for noninvasivehealth monitoring a reviewrdquo Journal of Medical Engineeringvol 2014 Article ID 381251 28 pages 2014

[18] J N M Kreider and L Hannapel ldquoElectrical impedanceplethysmography a physical and physiologic approach to pe-ripheral vascular studyrdquo Circulation vol 2 no 6 pp 811ndash8211950

[19] M Rehman B A Abu Izneid M Z Abdullah andM R Arshad ldquoAssessment of quality of fruits using impedancespectroscopyrdquo International Journal of Food Science andTechnology vol 46 no 6 pp 1303ndash1309 2011


Hindawiwwwhindawicom

International Journal of

Volume 2018

Zoology

Hindawiwwwhindawicom Volume 2018

Anatomy Research International

PeptidesInternational Journal of



Journal of Parasitology Research

GenomicsInternational Journal of


Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom

The Scientific World Journal

Volume 2018


BioinformaticsAdvances in

Marine BiologyJournal of



Neuroscience Journal


BioMed Research International

Cell BiologyInternational Journal of



Biochemistry Research International

ArchaeaHindawiwwwhindawicom Volume 2018


Genetics Research International


Advances in

Virolog y Stem Cells International



Enzyme Research



MicrobiologyHindawiwwwhindawicom

Nucleic AcidsJournal of

Volume 2018

Submit your manuscripts atwwwhindawicom

postharvest processing [4 5] MRI method requires sepa-ration of signal of water proton from that of fat proton whichis still a challenging task And both MRI and CT are limitedby the factors such as high cost and processing complexityHyperspectral imaging can effectively assess ripening de-grees [6] but it also suffers from constrains such as cost andprocessing erefore it is essential to expand currenttechnologies from different viewpoints On the contraryelectrical impedance spectroscopy (EIS) is a fast low costand nondestructive method which is found to offer insightinto plant physiology and physiological dynamics due toenvironmental impacts In this direction the EIS studieshave been conducted as a nondestructive evaluation methodto investigate the impedance spectrum variations and todetermine the ripening degree in the avocado

To develop an easily accessible and nondestructivemethod to understand mechanisms of ripening and to assessripening degree of the avocado the prospect of EIS tech-nique is explored in this paper is work on the avocadomainly focuses on the investigation of feasibility of EIS forassessment of the ripening degree nondestructively e restof the paper is organized as follows related past works onapplication of EIS technique is reviewed in the LiteratureReview section e next section introduces the theory ofbioimpedance and electrical impedance spectroscopy Lateron experimental methods and materials are illustratedAfter that the experimental and simulated results are pre-sented and discussed and finally paper is concluded in-cluding some future exploration directions

2 Literature Review

Impedance sensing technology especially EIS has emergeda new era into food quality and stability over the last decadesIt has been extensively used in the field of plant physiologyagriculture and food engineering for quality control andassessment of fruits and vegetables such as banana [7] Garutcitrus [8] kiwi [9] lettuce [10] nectarine [11] and straw-berry [12] In order to assess the freshness of banana EISinvestigation was performed during different ripening states[7] By attaching AgAgCl electrode and injecting a smallamount of current impedance responses are measured bythe 4294A impedance analyzer over a frequency range of50Hz to 1MHz e impedance magnitude phase anglereal part and imaginary part varied markedly with the al-teration of the ripening state

Gonzalez-Araiza et al [13] designed a nondestructivedevice to obtain the impedance spectrum of the wholestrawberry fruit and later on performed classification byutilizing corresponding equivalent circuit parameters (con-stant phase element CPE-P and Rinfinity) e study showedthat the strawberries at the highest stage of ripeness hadsignificantly lower constant phase element and Ro (related toextracellular) values compared to other strawberries

Neto et al [14] utilized EIS technique for the de-termination of the maturation degree of mangoes based onvariation of bulk resistance dependence with maturation offruits ey came up with this strategy to normalize bulkresistance by diameter to compensate the size variations of

samples ey demonstrated good agreement between vari-ation of electrical response and mechanical response of fruit

Montoya et al [15] investigated that electrical conduc-tivity could be a suitable factor for assessment of qualityduring ripening and cold storage ey found some re-semblances between electrical conductivity response andethylene productions curve ey defined a threshold ofconductivity (024 Sm) that indicates the limiting value forfruit stored at noninjurious temperatures and subsequentlytransferred to 20degC for marketing

Chowdhury et al [16] carried out EIS study on themandarin orange fruit during ripening in a spectrum be-tween 50Hz and 1MHz ey observed significant variationof the impedance phase angle real part and imaginary partof the impedance with different states of orange ripeninge study also demonstrated loss of weight of correspondingsamples with the progression of ripening states uselectrical sensing specially EIS technique was found to offerinsight into physiology of fruits and vegetables undergoingthe ripening process

3 Theory of Bioimpedance and ElectricalImpedance Spectroscopy

e plant body is a complex biological structure composedof tissues which are developed with cells suspended in ex-tracellular fluids (ECF) [17] Again cells are composed ofintracellular fluids (ICF) cell membrane (CM) and cell wall(CW) ECF ICF and CM are developed with differentmaterials and so exhibit distinguishable electrical attributese ECF and ICF act as electrolytes and provide a con-ducting path to applied alternating current [18] e CM isa protein-lipid-protein (P-L-P) structure and exhibits ca-pacitance to the current Consequently the overall responseof the biological tissues to an alternating electrical signalgenerates a complex sbioelectrical impedance Mathemati-cally the impedance Zang(θ) is calculated by dividing thevoltage (Vang(θ1)) measured by applied current (Iang(θ2)) as

Zang(θ) Vang θ1( 1113857

Iang θ2( 1113857 (1)

Bioimpedance is a complex quantity which varies withtissue composition and frequency of the applied signalerefore the frequency-dependent bioimpedance can berepresented as

Zb(ω) Re(Z(ω)) minus jIm(Z(ω)) Rb(ω)minus jXb(ω) (2)

where Rb(ω) and Xb(ω) represent the magnitude of the realpart of the complex Zb(ω) and the magnitude of theimaginary part of the complex Zb(ω) respectively

Bioimpedance is sensitive to the physiological status ofplant tissue Again as the response of bioimpedancechanges with frequency a multifrequency impedanceanalysis can offer better insight into plant physiology andbetter understanding to plant tissue status Electricalimpedance spectroscopy (EIS) is a multifrequency analy-sis for studying complex electrical impedance Z(ω)and its phase angle θ(ω) at different frequency points










minueξ

12u

Tu + C 1113944

i

l1ξl


(3)






(a)

(b)


Firm









05

1

15

2

25

3

Impe

danc

e mag

nitu

de in

ohm

times104

FirmBreaking

RipeOverripe

(a)


ndash70

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

Neg

ativ

e pha

se an

gle i

n de

gree

times104

FirmBreaking

RipeOverripe

(b)







times104

05

1

15

2

25

3

35

Impe

danc

e mag

nitu

de in

ohm

FirmBreaking

RipeOverripe

(a)


FirmBreaking

RipeOverripe

0

02

04

06

08

1

12

14

Neg

ativ

e pha

se an

gle i

n ra

dian

(b)

times104


FirmBreaking

RipeOverripe

04

06

08

1

12

14

16

18

2

22

Real

par

t of i

mpe

danc

e in

ohm

(c)

times104


FirmBreaking

RipeOverripe

0

05

1

15

2

25

3

35

Imag

inar

y pa

rt o

f im

peda

nce i

n oh

m

(d)








ndash15

ndash1

ndash05

0

05

1

PCA

2

times104

times104


FirmBreaking

RipeOverripe

0 2 4 6

(a)

times104

times104

ndash08

ndash06

ndash04

ndash02

0

02

04

06

FirmBreaking

RipeOverripe

PCA

2



0 2 4 6

(b)

times104

times104

ndash15

ndash1

ndash05

0

05

1

15

FirmBreaking

RipeOverripe

PCA1

PCA

2



(c)

times104

times104

ndash08ndash1


002040608

1

FirmBreaking

RipeOverripe

PCA1

PCA

2



(d)









6 Conclusion


Data Availability




References



Class 0 firm

Clas

s 0 f

irmClass 1 breaking

Clas

s 1 b

reak

ing

Class 2 ripe

Clas

s 2 r

ipe

Class 3 overripe

Clas

s 3 o

verr

ipe

Predicted label

8

8

8

6

4

2

2

2

10

10

10

0 0 0

0

0

0

0

0

0

0

0

True

labe

l



10

08

08

06

06

04




True

pos

itive

rate

02

020000


























Volume 2018

Zoology











Volume 2018

















Advances in




Enzyme Research





Volume 2018










minueξ

12u

Tu + C 1113944

i

l1ξl


(3)






(a)

(b)


Firm









05

1

15

2

25

3

Impe

danc

e mag

nitu

de in

ohm

times104

FirmBreaking

RipeOverripe

(a)


ndash70

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

Neg

ativ

e pha

se an

gle i

n de

gree

times104

FirmBreaking

RipeOverripe

(b)







times104

05

1

15

2

25

3

35

Impe

danc

e mag

nitu

de in

ohm

FirmBreaking

RipeOverripe

(a)


FirmBreaking

RipeOverripe

0

02

04

06

08

1

12

14

Neg

ativ

e pha

se an

gle i

n ra

dian

(b)

times104


FirmBreaking

RipeOverripe

04

06

08

1

12

14

16

18

2

22

Real

par

t of i

mpe

danc

e in

ohm

(c)

times104


FirmBreaking

RipeOverripe

0

05

1

15

2

25

3

35

Imag

inar

y pa

rt o

f im

peda

nce i

n oh

m

(d)








ndash15

ndash1

ndash05

0

05

1

PCA

2

times104

times104


FirmBreaking

RipeOverripe

0 2 4 6

(a)

times104

times104

ndash08

ndash06

ndash04

ndash02

0

02

04

06

FirmBreaking

RipeOverripe

PCA

2



0 2 4 6

(b)

times104

times104

ndash15

ndash1

ndash05

0

05

1

15

FirmBreaking

RipeOverripe

PCA1

PCA

2



(c)

times104

times104

ndash08ndash1


002040608

1

FirmBreaking

RipeOverripe

PCA1

PCA

2



(d)









6 Conclusion


Data Availability




References



Class 0 firm

Clas

s 0 f

irmClass 1 breaking

Clas

s 1 b

reak

ing

Class 2 ripe

Clas

s 2 r

ipe

Class 3 overripe

Clas

s 3 o

verr

ipe

Predicted label

8

8

8

6

4

2

2

2

10

10

10

0 0 0

0

0

0

0

0

0

0

0

True

labe

l



10

08

08

06

06

04




True

pos

itive

rate

02

020000


























Volume 2018

Zoology











Volume 2018

















Advances in




Enzyme Research





Volume 2018


(a)

(b)


Firm









05

1

15

2

25

3

Impe

danc

e mag

nitu

de in

ohm

times104

FirmBreaking

RipeOverripe

(a)


ndash70

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

Neg

ativ

e pha

se an

gle i

n de

gree

times104

FirmBreaking

RipeOverripe

(b)







times104

05

1

15

2

25

3

35

Impe

danc

e mag

nitu

de in

ohm

FirmBreaking

RipeOverripe

(a)


FirmBreaking

RipeOverripe

0

02

04

06

08

1

12

14

Neg

ativ

e pha

se an

gle i

n ra

dian

(b)

times104


FirmBreaking

RipeOverripe

04

06

08

1

12

14

16

18

2

22

Real

par

t of i

mpe

danc

e in

ohm

(c)

times104


FirmBreaking

RipeOverripe

0

05

1

15

2

25

3

35

Imag

inar

y pa

rt o

f im

peda

nce i

n oh

m

(d)








ndash15

ndash1

ndash05

0

05

1

PCA

2

times104

times104


FirmBreaking

RipeOverripe

0 2 4 6

(a)

times104

times104

ndash08

ndash06

ndash04

ndash02

0

02

04

06

FirmBreaking

RipeOverripe

PCA

2



0 2 4 6

(b)

times104

times104

ndash15

ndash1

ndash05

0

05

1

15

FirmBreaking

RipeOverripe

PCA1

PCA

2



(c)

times104

times104

ndash08ndash1


002040608

1

FirmBreaking

RipeOverripe

PCA1

PCA

2



(d)









6 Conclusion


Data Availability




References



Class 0 firm

Clas

s 0 f

irmClass 1 breaking

Clas

s 1 b

reak

ing

Class 2 ripe

Clas

s 2 r

ipe

Class 3 overripe

Clas

s 3 o

verr

ipe

Predicted label

8

8

8

6

4

2

2

2

10

10

10

0 0 0

0

0

0

0

0

0

0

0

True

labe

l



10

08

08

06

06

04




True

pos

itive

rate

02

020000


























Volume 2018

Zoology











Volume 2018

















Advances in




Enzyme Research





Volume 2018






05

1

15

2

25

3

Impe

danc

e mag

nitu

de in

ohm

times104

FirmBreaking

RipeOverripe

(a)


ndash70

ndash60

ndash50

ndash40

ndash30

ndash20

ndash10

Neg

ativ

e pha

se an

gle i

n de

gree

times104

FirmBreaking

RipeOverripe

(b)







times104

05

1

15

2

25

3

35

Impe

danc

e mag

nitu

de in

ohm

FirmBreaking

RipeOverripe

(a)


FirmBreaking

RipeOverripe

0

02

04

06

08

1

12

14

Neg

ativ

e pha

se an

gle i

n ra

dian

(b)

times104


FirmBreaking

RipeOverripe

04

06

08

1

12

14

16

18

2

22

Real

par

t of i

mpe

danc

e in

ohm

(c)

times104


FirmBreaking

RipeOverripe

0

05

1

15

2

25

3

35

Imag

inar

y pa

rt o

f im

peda

nce i

n oh

m

(d)








ndash15

ndash1

ndash05

0

05

1

PCA

2

times104

times104


FirmBreaking

RipeOverripe

0 2 4 6

(a)

times104

times104

ndash08

ndash06

ndash04

ndash02

0

02

04

06

FirmBreaking

RipeOverripe

PCA

2



0 2 4 6

(b)

times104

times104

ndash15

ndash1

ndash05

0

05

1

15

FirmBreaking

RipeOverripe

PCA1

PCA

2



(c)

times104

times104

ndash08ndash1


002040608

1

FirmBreaking

RipeOverripe

PCA1

PCA

2



(d)









6 Conclusion


Data Availability




References



Class 0 firm

Clas

s 0 f

irmClass 1 breaking

Clas

s 1 b

reak

ing

Class 2 ripe

Clas

s 2 r

ipe

Class 3 overripe

Clas

s 3 o

verr

ipe

Predicted label

8

8

8

6

4

2

2

2

10

10

10

0 0 0

0

0

0

0

0

0

0

0

True

labe

l



10

08

08

06

06

04




True

pos

itive

rate

02

020000


























Volume 2018

Zoology











Volume 2018

















Advances in




Enzyme Research





Volume 2018






times104

05

1

15

2

25

3

35

Impe

danc

e mag

nitu

de in

ohm

FirmBreaking

RipeOverripe

(a)


FirmBreaking

RipeOverripe

0

02

04

06

08

1

12

14

Neg

ativ

e pha

se an

gle i

n ra

dian

(b)

times104


FirmBreaking

RipeOverripe

04

06

08

1

12

14

16

18

2

22

Real

par

t of i

mpe

danc

e in

ohm

(c)

times104


FirmBreaking

RipeOverripe

0

05

1

15

2

25

3

35

Imag

inar

y pa

rt o

f im

peda

nce i

n oh

m

(d)








ndash15

ndash1

ndash05

0

05

1

PCA

2

times104

times104


FirmBreaking

RipeOverripe

0 2 4 6

(a)

times104

times104

ndash08

ndash06

ndash04

ndash02

0

02

04

06

FirmBreaking

RipeOverripe

PCA

2



0 2 4 6

(b)

times104

times104

ndash15

ndash1

ndash05

0

05

1

15

FirmBreaking

RipeOverripe

PCA1

PCA

2



(c)

times104

times104

ndash08ndash1


002040608

1

FirmBreaking

RipeOverripe

PCA1

PCA

2



(d)









6 Conclusion


Data Availability




References



Class 0 firm

Clas

s 0 f

irmClass 1 breaking

Clas

s 1 b

reak

ing

Class 2 ripe

Clas

s 2 r

ipe

Class 3 overripe

Clas

s 3 o

verr

ipe

Predicted label

8

8

8

6

4

2

2

2

10

10

10

0 0 0

0

0

0

0

0

0

0

0

True

labe

l



10

08

08

06

06

04




True

pos

itive

rate

02

020000


























Volume 2018

Zoology











Volume 2018

















Advances in




Enzyme Research





Volume 2018







ndash15

ndash1

ndash05

0

05

1

PCA

2

times104

times104


FirmBreaking

RipeOverripe

0 2 4 6

(a)

times104

times104

ndash08

ndash06

ndash04

ndash02

0

02

04

06

FirmBreaking

RipeOverripe

PCA

2



0 2 4 6

(b)

times104

times104

ndash15

ndash1

ndash05

0

05

1

15

FirmBreaking

RipeOverripe

PCA1

PCA

2



(c)

times104

times104

ndash08ndash1


002040608

1

FirmBreaking

RipeOverripe

PCA1

PCA

2



(d)









6 Conclusion


Data Availability




References



Class 0 firm

Clas

s 0 f

irmClass 1 breaking

Clas

s 1 b

reak

ing

Class 2 ripe

Clas

s 2 r

ipe

Class 3 overripe

Clas

s 3 o

verr

ipe

Predicted label

8

8

8

6

4

2

2

2

10

10

10

0 0 0

0

0

0

0

0

0

0

0

True

labe

l



10

08

08

06

06

04




True

pos

itive

rate

02

020000


























Volume 2018

Zoology











Volume 2018

















Advances in




Enzyme Research





Volume 2018





6 Conclusion


Data Availability




References



Class 0 firm

Clas

s 0 f

irmClass 1 breaking

Clas

s 1 b

reak

ing

Class 2 ripe

Clas

s 2 r

ipe

Class 3 overripe

Clas

s 3 o

verr

ipe

Predicted label

8

8

8

6

4

2

2

2

10

10

10

0 0 0

0

0

0

0

0

0

0

0

True

labe

l



10

08

08

06

06

04




True

pos

itive

rate

02

020000


























Volume 2018

Zoology











Volume 2018

















Advances in




Enzyme Research





Volume 2018

























Volume 2018

Zoology











Volume 2018

















Advances in




Enzyme Research





Volume 2018


research articledownloads.hindawi.com/journals/jfq/2018/4706147.pdfresearch article assessment of...

Documents