Journal of Food Engineering 128 (2014) 140–145

Contents lists available at ScienceDirect

Journal of Food Engineering

journal homepage: www.elsevier .com/locate / j foodeng

Investigation of the application of digital colour imaging to assessthe mixture quality of binary food powder mixes

0260-8774/$ - see front matter � 2014 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.jfoodeng.2013.12.013

⇑ Corresponding author. Tel.: +46 704922723.E-mail address: lilia.ahrne@sik.se (L. Ahrné).

Pooja Shenoy a,b, Fredrik Innings c, Torbjörn Lilliebjelke d, Caroline Jonsson d, John Fitzpatrick b,Lilia Ahrné a,⇑a SIK, The Swedish Institute for Food and Biotechnology, Gothenburg, Swedenb Department of Process and Chemical Engineering, University College Cork, Irelandc Tetra Pak Processing Systems, Research Technology, Lund, Swedend Santa Maria AB, Mölndal, Sweden

a r t i c l e i n f o a b s t r a c t

Article history:Received 16 August 2013Received in revised form 9 December 2013Accepted 12 December 2013Available online 21 December 2013

Keywords:Food powdersMixingMixture qualityColour imaging

Digital colour imaging (DCI) was applied in this study as a novel approach for assessing the mixture qual-ity of binary food powder mixes. Three different binary powder mixes with different coloured ingredients[salt, paprika, black pepper and onion] were investigated using a commercially available system calledDigiEye. The coordinates of CIELAB colour space were used to describe the colour of the samples. Thesample colour variance was used as a measure of mixture quality. The results showed that DCI has poten-tial for assessing the mixture quality of binary food powder mixes, provided that colour differencebetween the powders can be measured. The ability to assess mixture quality decreases as the colour dif-ference between the components in the mix decreases. Furthermore, scale of scrutiny and compositionalso influence the capability of the method.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction methods to evaluate the powder mixture homogeneity include

Powder mixing is a complex operation that requires knowledgeof powder properties such as powder size, shape, bulk density, par-ticle density, mixer design and mixing conditions. Mixing opera-tions are undertaken to obtain the best homogeneity in themulti-component mixture and also to ensure consistent productperformance (Aissa et al., 2011). A homogeneous mixture is onein which the composition of all constituents are uniform through-out the whole mixture and it is important to be able to assess thestate of homogeneity of the mixture (Fan et al., 1970; Aissa et al.,2010).

In many food industries, there is little or no objective assess-ment of food powder mixture quality and there is a need in indus-try for quick and user-friendly techniques that are potentiallyfaster for analyzing mixture homogeneity, such as near infra-redspectroscopy (Kehlenbeck, 2011) and image analysis (Aissa et al.,2011). Image analysis techniques are becoming increasingly popu-lar due to the speed of analysis, lower costs and also because theyare simpler to use (Aissa et al., 2010). They can be used as a com-plementary method with the traditional methods e.g. along withparticle sizing methods where it acts as a means to validate thedata obtained (Boschetto and Giordano, 2012). Traditional

use of thief probes for sampling followed by estimation of particlesize distribution (Boschetto and Giordano, 2012) or measurementof conductivity when salt is one of the ingredients or applicationof UV spectroscopy when active ingredients are used (Muzzioet al., 2003). Thief probe sampling leads to destruction of the origi-nal state of the powder bed as compared to image analysis tech-niques that are non-destructive and faster to use in on-linesystems (Muzzio et al., 2003; Berthiaux et al., 2006). Howeverone drawback with image analysis is that lighting conditionsmay not be stable and thus require background correction on eachimage to avoid errors during analysis (Berthiaux et al., 2006; Le Co-ent et al., 2005; Muerza et al., 2002).

In the spice industries where food powder mixtures are pro-duced with many ingredients that vary in shape, size, colour andtexture (Barbosa-Cánovas et al., 2005), colour and the visualappearance is of prime importance to appeal to the customer. Inindustry, manual assessment of mixture quality is sometimesundertaken by visual inspection of the uniformity of colour withinthe mix. In fresh produce industry, such as in harvesting and pur-chase of fruits and vegetables, visual assessment of colour plays animportant role in determining the value of the product and tojudge if it is ripe enough (Ji et al., 2013; Jha and Matsouka,2000). However, visual assessment in the food industry needs tobe standardized since this can vary depending on the personassessing the sample, the illumination conditions and also the an-gle of observance. Thus, there is a need to investigate the

Nomenclature

DCI digital colour imagingL�, a�, b� CIELAB colour coordinatesL�standard; a

�standard; b

�standard CIELAB colour coordinates for white

spaceDE colour parameter relative to white

space

N number of squares in the imagel average DE for well mixed sampleVariance DE variance in DE values for N squaresVariance DENM variance in DE for non-mixed powders

Fig. 1. DigiEye image of paprika powder partitioned into 25 squares (grey-scale).

P. Shenoy et al. / Journal of Food Engineering 128 (2014) 140–145 141

application of objective assessment methods which are quick andreliable to assess the mixture quality in industry. Colour imagingcan help to track every particle since the colour of each pixel hasspecific values in the Red Green Blue (RGB) coordinates (Aissaet al., 2010). Previous studies that have applied colour assessmentmethod have been to assess the ripeness of banana (Ji et al., 2013),the firmness of mangoes (Jha et al., 2006) and to correlate the sur-face gloss with weight of eggplant during its storage (Jha et al.,2002). Hunter L, a, and b values and maturity index were modeledto evaluate maturity of mango by non-destructive means (Jha et al.,2007). The objective of this study was to explore the application ofdigital colour imaging (DCI) for assessing food powder mixturequality and to highlight any problems and limitations associatedwith it. The DCI system used in the study was DigiEye which hasa high resolution camera where the lighting conditions for imagingare controlled.

2. Materials and methods

2.1. Powders and binary mixes

The food powders used were paprika, salt, black pepper andonion. All the powders were obtained from Santa Maria, Gothen-burg, Sweden. The binary powders mixes used were:

� Paprika–Salt� Black pepper–Salt� Onion–Salt

Three composition recipes (by mass) were mixed for each of thebinary mixes, that is 30:70; 50:50 and 70:30. The volume meandiameter, D [4, 3], for salt, paprika, black pepper and onion pow-ders is 454 lm, 252 lm, 369 lm and 65 lm, respectively.

2.2. DigiEye and its operation

The digital colour imaging system (DigiEye), sourced from Veri-Vide Ltd. UK is a complete non-contact colour imaging and mea-surement system. This equipment consists of a digital cameraNikon D90 with an image quality of 12.3 megapixels and a samplechamber with fluorescent D65 illuminant to provide controlledlight conditions inside the cabinet. This helps the collection ofsame quality high resolution images which do not require back-ground lighting correction. It has the ability to measure colour atmany points of the sample region. The colour measurement datais reported as colorimetric values such as XYZ and CIELAB and spec-tral reflectance between the range of 400–700 nm with an intervalof 10 nm. The DigiEye has been used for colour assessment in boththe food industry and textile industry (VeriVide Ltd, 2010; DigiEyeUser Guide, 2007).

2.3. Mixing procedure and assessment of mixture quality

The binary powder mixes were prepared by weighing the re-quired amount of ingredients in a plastic transparent bag kept on

a calibrated weighing scale. In each trial the total amount of binarymix prepared was 100 g. The bag was tied and then shaken manu-ally by rotating the bag right and left (to mimic a tumble mixertype of movement). The extent of mixing equalled the number oftimes the bag was shaken from right to left. Trials were undertakenwhereby powder samples were shaken for a different number oftimes in order to mimic a progression in mixing. In some trials,the powders were shaken many times, approximately 10–15 times,to obtain ‘‘well-mixed’’ powders or what were perceived to be wellmixed to the human eye.

At the end of mixing the powder mixture was poured onto acustom made transparent plexiglass dish [100 � 100 � 10 mm]and levelled using a ruler. The sample was placed inside the Digi-Eye chamber and an image of the mixture in the dish was mea-sured by the DigiEye camera. The image was virtually dividedinto 25 or 100 small squares, as illustrated in Fig. 1. For each squarethe CIELAB colour coordinates L�, a� and b� were measured and DE(the colour value relative to white space) was calculated from Eq.(1):

DE ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiL� � L�standard

� �2 þ a� � a�standard

� �2 þ b� � b�standard

� �2q

ð1Þ

where L�standard; a�standard and b�standard are the values measured for

white space.The variance in DE (Variance DE) was used as a measure of mix-

ture quality of the powder mix and this was calculated from Eq.(2):

Variance DE ¼PN

i¼1 DEi � lð Þ2

Nð2Þ

where l is the average DE for well-mixed sample and N is the num-ber of squares in the DigiEye image.

The variance in DE for a non-mixed powder mix at the begin-ning of each trial (Variance DENM) was calculated from Eq. (3).Eq. (2) was then applied thereafter.

Variance DENM ¼ NP DEP � lð Þ2 þ NS DES � lð Þ2 ð3Þ

where NP and NS are the mass fractions of the ingredient powder(paprika or black pepper or onion) and salt, respectively and DEP

0

500

1000

1500

2000

2500

3000

0 1 2 3 4 5 6 7 8 9

Extent of mixing

Varia

nce

E

Paprika-Salt Pepper-Salt Onion-Salt

0.01

0.1

1

10

100

1000

10000

0 1 2 3 4 5 6 7 8 9Extent of mixing

Varia

nce

E

Paprika-Salt Pepper-Salt Onion-Salt

(a)

(b)

Fig. 2. Variance DE [25 squares] for the three binary mixes as a function of extent ofmixing for 70% salt concentration: (a) non-log and (b) log plot.

142 P. Shenoy et al. / Journal of Food Engineering 128 (2014) 140–145

and DES are the DE values for the ingredient powder and saltrespectively (note: salt is used in all the binary mixes).

2.4. Repeatability studies

Triplicate trials were performed for all the well mixed samplesand good repeatability was obtained with standard deviations inDE of less than 0.8, 0.1 and 0.3 for paprika–salt, pepper–salt andonion–salt compositions, respectively. In the case of partiallymixed samples, repeatability studies were not performed as itwas not possible to replicate partial mixing by shaking.

3. Results and discussion

3.1. Powder colour values and colour differences between powders

Digital colour imaging (DCI) methods use differences in colourvalues to assess mixture quality, thus there must be a differencein a colour value, such as DE, of the individual powders to bemixed. Table 1 presents CIELAB colour values obtained by DCI sys-tem (DigiEye) for the individual powders, the differences betweenthem and also the colour values for well-mixed binary mixes. Forall the binary mixes, there are differences in DE. The largest isfor paprika–salt at 73 and the smallest is for onion–salt at 5.6, thusit would be expected that the mixture quality for onion–salt wouldbe the most difficult to assess, as the onion and salt have very sim-ilar colours.

The well-mixed paprika–salt mixes have DE colour values muchcloser to the paprika value than the salt value which shows thatpaprika is having a dominant effect on the DE value. DE progres-sively increases as the paprika content increases as expected. Like-wise, a similar trend was obtained for the pepper–salt mixes. Thewell-mixed onion–salt mixes also had DE colour values close tothe onion, but there was no trend with increasing onion composi-tion for the range presented and this suggests that it may not bepossible to assess mixture quality of this mix using DCI.

3.2. Mixture quality assessment using digital colour imaging

Data from the colour measurements for the three binary mixesas a function of mixing extent (or number of shakes) are presentedfor the 70% and 30% salt compositions in Figs. 2 and 3, respectively.DE was measured for each of 25 squares [as illustrated in Fig. 1]and variance DE was calculated using Eq. (2). The data are pre-sented both as non-log and log plots for greater clarity. Fig. 2ashows that the variance DE decreases rapidly upon shaking andthat it is highest for paprika–salt and lowest for onion–salt, as ex-

Table 1Colour values (CIELAB) and differences in colour (relative to white values) forindividual powders, and colour values for well-mixed binary mixes.

Name of sample DE a� b� L�

Salt 4.70 �0.7 0.8 96.7Paprika 77.7 35.2 39.0 33.7Black pepper 51.2 5.0 17.2 43.7Onion powder 10.3 �0.2 11.9 92.3Difference [paprika–salt] 73 35.9 38.2 �63.0Difference [black pepper–salt] 46.4 5.7 16.4 �53.0Difference [onion–salt] 5.6 0.4 11.1 �4.4Well mixed paprika–salt [30:70] 64.9 29.7 29.3 41.4Well mixed paprika–salt [50:50] 72.0 32.9 32.8 36.0Well mixed paprika–salt [70:30] 75.3 33.7 35.0 33.7Well mixed black pepper–salt [30:70] 42.4 3.7 13.2 51.5Well mixed black pepper–salt [50:50] 47.7 4.3 14.5 46.4Well mixed black pepper–salt [70:30] 50.3 4.4 15.5 44.0Well mixed onion powder–salt [30:70] 10.4 0.1 12.1 92.0Well mixed onion powder–salt [50:50] 10.1 �0.2 11.6 91.4Well mixed onion powder–salt [70:30] 10.4 �0.2 11.9 93.6

pected from the values presented in Table 1. Data scatter wasquantitatively estimated using the root mean square error of thelinear regression of a log plot and these values are presented inTable 2. Fig. 2b and Table 2 show there is more scatter in theonion–salt data, suggesting that DCI may have greater difficultyin distinguishing colour differences for these components. Thereis less scatter for the other 2 mixes and there is a gradual reductionin variance DE as mixing progresses.

Fig. 3a shows the variance DE values for all mixes at 30% salt area lot lower than their corresponding values for 70% salt (Fig. 2).Furthermore, Fig. 3b and Table 2 show that the data for onion–saltis highly scattered at 30% salt content and suggests that DCI isunable to distinguish colour differences at this salt content.Overall, DCI has less capability for assessing mixture quality atthe lower salt content.

In Figs. 2 and 3 it is seen that for the 70% and 30% salt mixes thevariance DE decreases with mixing extent for the paprika–salt andpepper–salt mixes as they approach a well-mixed state. However,there were lower values for variance DE at lower salt content(Fig. 3) which suggests that composition may influence the appli-cability of DCI. These data show that there is potential for usingDCI for assessing the mixture quality of these powders. It alsoshows that the approach may be limited if the difference in DE istoo small.

3.3. Effect of powder composition on colour measurement

For a binary mix containing powders with different colours, onewould expect to have colour variations in well mixed binary mixes

0

200

400

600

800

1000

1200

1400

1600

Extent of mixing

Varia

nce

E Paprika-Salt Pepper-Salt Onion-Salt

0.01

0.1

1

10

100

1000

10000

0 1 2 3 4 5 6

0 1 2 3 4 5 6

Extent of mixing

Varia

nce

E

Paprika-Salt Pepper-Salt Onion-Salt

(a)

(b)

Fig. 3. Variance DE [25 squares] for the three binary mixes as a function of extent ofmixing for 30% salt concentration: (a) non-log and (b) log plot.

Table 2Estimation of data scatter using the root mean square error of the linear regression ofthe logarithmic plots in Figs. 2b, 3b, 5b and 6b [The first data point (i.e. non-mixedpowders) in each figure was omitted in the regression analysis].

25 squares 100 squares

70% Salt 30% Salt 70% Salt 30% Salt

Paprika–salt 0.76 0.51 0.97 0.93Pepper–salt 0.82 0.89 0.9 0.97Onion–salt 0.6 0.06 0.86 0.87

0

10

20

30

40

50

60

70

80

90

Paprika, Pepper or Onion mass fraction

Paprika/SaltPepper/SaltOnion/Salt

0

2

4

6

8

10

12

0 0.2 0.4 0.6 0.8 1

0 0.2 0.4 0.6 0.8 1

Onion mass fraction

(a)

(b)

ΔE

ΔE

Fig. 4. Effect of composition on DE of binary mixes: (a) all mixes and (b) onion–salt.

P. Shenoy et al. / Journal of Food Engineering 128 (2014) 140–145 143

containing different fractions of the components. Consequently, itmight be expected that DCI should be able to differentiate betweenwell-mixed mixes with different compositions. Trials were under-taken with binary mixes at 10% intervals between 0% and 100% saltfor each of the binary mixes. Each mix was shaken several timesuntil it was well-mixed. DCI was used to measure DE for eachmix. Fig. 4 shows how DE varies with % salt for each of the binarymixes.

For paprika–salt, there is an increase in DE over the entire con-centration range; however the rate of increase in DE is much great-er below 50% paprika. Consequently, the increase in DE becomesless sensitive as the paprika content is increased above 50%. Visualinspection of the well-mixed samples also showed that it was dif-ficult to distinguish between paprika–salt mixes in the higher pa-prika concentration range. For pepper–salt, there was a gradualincrease in DE up to about 80% pepper, after which this relation-ship did not exist which suggests that Digi-eye may not be able

to detect concentration differences above 80%. The greatest rateof increase was below 50%. For onion–salt, there was only an in-crease in DE up to about 30% onion, after which the data was scat-tered. This suggests that DCI may not be able to or may have majordifficulty in assessing the mixture quality of onion–salt mixes be-cause the difference in DE is too small. This is in agreement withdata presented in Figs. 2 and 3.

In Fig. 3, the variance DE of the pepper–salt mixes decreasedmore rapidly during mixing than the paprika–salt which may sug-gest that very good mixture quality was obtained much more rap-idly in the pepper–salt trial. Furthermore, comparing data forpepper–salt and paprika–salt in Figs. 2 and 3 would suggest thatvery good mixture quality was obtained in the 30% salt trials. How-ever, this may be misleading because DCI has greater variance DEvalues for paprika–salt than pepper–salt and it has reduced detec-tion capability for colour differences in the higher 70% paprika andpepper contents. So, the much more rapid improvement in mixturequality for both paprika–salt and pepper–salt may be erroneousbecause the results plotted may be due to DCI’s lower colour differ-ence detection capability above the 50% concentration range.

3.4. Effect of sample size on mixture quality assessment using DCI

A consequence of Fig. 4 is that DCI may have major difficulty inassessing the mixture quality of, for example, pepper–salt mixeswith high pepper content. This is because any colour deviationsand corresponding variance that occur in the high pepper concen-tration range may not be detected. One approach to try and over-come this problem is to reduce the sample size. In powder

1

10

100

1000

10000

Varia

nce

E

Paprika-Salt Pepper-Salt Onion-Salt

0.01

200

400

600

800

1000

1200

1400

1600

0 1 2 3 4 5 6

Extent of mixing

Varia

nce

E

Paprika-Salt Pepper-Salt Onion-Salt

(a)

(b)

144 P. Shenoy et al. / Journal of Food Engineering 128 (2014) 140–145

mixing, variance depends on sample size with variance increasingwith smaller sample size (Fitzpatrick, 2009). In the DCI procedure,the sample size is the size of the squares, thus reducing the size ofthe squares could potentially increase the variance. Reducing sam-ple size will increase the variation in component concentrationsbetween individual squares. This will increase variance DE, pro-vided that some of the squares have concentrations within the con-centration range where DCI has sensitivity.

Consequently, trials were carried out with the binary mixeswith a reduced sample size by increasing the number of squaresin the colour image from 25 to 100, that is, a fourfold reductionin sample size. This is also represented by a corresponding reduc-tion in the number of pixels in a square. The data from these trialsare presented in Figs. 5 and 6 for 70% and 30% salt contents, respec-tively. Fig. 5 shows that higher variance DE values were obtainedfor smaller sample sizes (in comparison with Fig. 2) and therewas a much smoother reduction in variance DE during mixing withlittle data scatter (Table 2), even for the onion–salt mix. Fig. 6 alsoshows higher variance DE values in comparison to Fig. 3 and thereis also a smooth gradual reduction in variance DE with little scatter(Table 2). For onion–salt, the variance DE rapidly reduces to a verylow value and there is little difference between the values overmost of the range. This is signifying that DCI may not be able to as-sess mixture quality for the lower salt content of 30%.

Fig. 5 shows that DCI appears to be able to assess the mixturequality of the onion–salt mix at 30% onion content, even thoughFig. 4b would suggest it would have difficulty at this content. How-ever, at the sample size used in Fig. 5 there will be variation in

0.01

500

1000

1500

2000

2500

3000

Extent of mixing

Varia

nce

E

Paprika-Salt Pepper-Salt Onion-Salt

0.01

0.1

1

10

100

1000

10000

0 1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9Extent of mixing

Varia

nce

E

Paprika-Salt Pepper-Salt Onion-Salt

(a)

(b)

Fig. 5. Variance DE [100 squares] for the three binary mixes as a function of extentof mixing for 70% salt concentration: (a) non-log and (b) log plot.

0.01

0.1

Extent of mixing0 1 2 3 4 5 6

Fig. 6. Variance DE [100 squares] for the three binary mixes as a function of extentof mixing for 30% salt concentration: (a) non-log and (b) log plot.

onion content amongst the samples, thus some of the samplesmust have onion contents in the lower onion content range whereDCI can distinguish colour differences. At 70% onion content, theprobability of this occurring is a lot less and thus the low variancevalues signify the lack of colour difference detection capability inthis trial (Fig. 6). Overall, reducing the sample size improved theability for assessing mixture quality, even for the onion–salt pow-der mix.

3.5. Comparing mixture quality of different mixes using DCI

The mixture quality of different mixes cannot be comparedbased on comparing the values of variance DE. For example,Fig. 5 shows that the variance DE for onion–salt is lower than thatof paprika–salt. This does not signify that there is a superior mix-ture quality being achieved in the onion–salt mix. The reason forthis is that it is not only mixture quality that influences the valueof variance DE; the difference in DE between the two componentpowders also has a major influence. This difference is much lowerfor onion–salt than paprika–salt and this is most likely the domi-nant reason why variance DE is lower for the onion–salt mixes.

3.6. Application of DCI and variance DE in quality control

DCI has potential for assessing the mixture quality of food pow-der mixes as a quality control tool. This is provided that there aresufficient colour differences between the component powders. Inquality control, a representative sample would be taken and the

P. Shenoy et al. / Journal of Food Engineering 128 (2014) 140–145 145

variance DE would be measured. This measured variance DE wouldneed to be compared to a limit variance DE below which the mea-sured variance DE would be deemed to represent satisfactory mix-ture quality. A key aspect in this comparison is determination ofthe limit variance DE. This is a value that a processor would haveto determine them self, for example by conducting a number ofexperimental trials. It is important to keep in mind that the samplesize or number of pixels in the sample will influence the value ofvariance DE, thus this should be noted and kept constant. Further-more cognisance must be given to other factors that could influ-ence variability in the evaluation of variance DE, such as batch tobatch variability in the colour of component powders and repeat-ability of the procedures used to evaluate variance DE.

4. Conclusions

Digital colour imaging (DCI) has potential for assessing the mix-ture quality of binary food powder mixes, provided that a colourdifference can be measured between the powders. The greaterthe colour difference, the greater the potential for the DCI basedapproach. Even for powders with different colours, DCI may havedifficulty in distinguishing between well-mixed samples with dif-ferent compositions, however it is critical that there is a range ofcompositions in which DCI can distinguish, otherwise the tech-nique will not work. Reducing the sample size, by reducing squaresize in the images or reducing the number of pixels in a square, canhelp improve the colour variance approach for assessing mixturequality.

Acknowledgement

This work was supported by the EU seventh frameworkprogramme through the ‘‘PowTech’’ Marie Curie Initial TrainingNetwork (Project No. EU FP7-PEOPLE-2010-ITN-264722).

References

Aissa, A.A., Duschesne, C., Rodrigue, D., 2010. Polymer powders mixing part II:multi-component mixing dynamics using RGB color analysis. Chem. Eng. Sci. 65,3729–3738.

Aissa, A.A., Duschesne, C., Rodrigue, D., 2011. Effect of friction coefficient anddensity on mixing particles in the rolling regime. Powder Technol. 212, 340–347.

Barbosa-Cánovas, G.V., Ortega-Rivas, E., Juliano, P., Yan, H., 2005. Particle properties.In: Food Powders – Physical Properties, Processing and Functionality. KluwerAcademic/Plenum Publishers, New York.

Berthiaux, H., Mosorov, V., Tomczak, L., Gatumel, C., Demeyre, J.F., 2006. Principalcomponent analysis for characterising homogeneity in powder mixing usingimage processing techniques. Chem. Eng. Process. 45, 397–403.

Boschetto, A., Giordano, V., 2012. Powder sampling and characterization by digitalimage analysis. Measurement 45, 1023–1038.

DigiEye User Guide, version 2.43, 2007. Unpublished Internal Document.Fan, L.T., Chen, S.J., Watson, C.A., 1970. Annual review solids mixing. Ind. Eng. Chem.

62, 53–69.Fitzpatrick, J.J., 2009. Particulate and powder mixing. In: Cullen, P.J. (Ed.), In Food

Mixing – Principles and Applications. Wiley Blackwell, Chichester, UK.Jha, S.N., Matsouka, T., 2000. Non-destructive techniques for quality evaluation of

intact fruits and vegetables. Food Sci. Technol. Res. 6 (4), 248–251.Jha, S.N., Matsouka, T., Miyauchi, K., 2002. Surface gloss and weight of eggplant

during storage. Biosyst. Eng. 81 (4), 407–412.Jha, S.N., Kingsly, A.R.P., Chopra, S., 2006. Non-destructive determination of firmness

and yellowness of mango during growth and storage using visual spectroscopy.Biosyst. Eng. 94 (3), 397–402.

Jha, S.N., Chopra, S., Kingsly, A.R.P., 2007. Modeling of color values fornondestructive evaluation of maturity of mango. J. Food Eng. 78, 22–26.

Ji, W., Koutsidis, G., Luo, R., Hutchings, J., Akhtar, M., Megias, F., Butterworth, M.,2013. Color Res. Appl. 38, 364–374.

Kehlenbeck, V., 2011. Use of near infrared spectroscopy for in- and off-lineperformance determination of continuous and batch powder mixers:opportunities and challenges. Procedia Food Sci. 1, 2015–2022.

Le Coent, A.L., Rivoire, A., Briancon, S., Lieto, J., 2005. An original image-processingtechnique for obtaining the mixing time: the box-counting with erosionsmethod. Powder Technol. 152, 62–71.

Muerza, S., Berthiaux, H., Massol-Chaudeur, S., Thomas, G., 2002. A dynamic studyof static mixing using on-line image analysis. Powder Technol. 128, 195–204.

Muzzio, F.J., Goodridge, C.L., Alexander, A., Arratia, P., Yang, H., Sudah, O., Mergen,G., 2003. Sampling and characterization of pharmaceutical powders andgranular blends. Int. J. Pharm. 250, 51–64.

VeriVide Ltd., 2010. The DigiEye System, Unpublished Internal Document.