Examples of images captured with the system for a honey sample are showm in Figure 5.3. An image obtained with polychromatic white from the VDU screen to provide transmission information is shown in Figure 5.3a, whilst an image with polarised white light from the VDU is showm in Figure
FIGURE 5.2 PC/webcam-based chromatic system for monitoring honey samples, (a) View of overall system, (b) Schematic diagram (Sufian. 2014).
5.3b and an image with fluorescence produced by the honey sample with the light from the LEDs is shown in Figure 5.3c.
R, G. В values were obtained for chosen test X(w) = R(w), G(w) or B(w) and reference X(ref) = R(ref), G(ref) or B(ref) areas in such images (e.g„ Figure 5.3a) for transmission, polarisation and fluorescence illumination. A correction factor X(cf) was applied to accommodate test-to-test variations, yielding test result X(w)s.
where X(ref)set is the chosen set values of X(ref).
FIGURE 5.3 Examples of images of a honey sample and background VDU screen under different optical conditions, (a) Optical transmission, (b) Polarised light, (c) Optical fluorescence (Sufian, 2014).
In addition, changing ambient conditions were accommodated via X(A) by placing a black card behind an empty cuvette and extracting RGB values in a similar manner to the test samples. The test results X(w)s were normalised according to samples corresponding to the maximum X(w)smax and minimum X(w)smin levels of the measurement ranges.
The previous procedures were applied to each of the optical signals, that is, transmission, polarisation and fluorescence.
For transmission, X(w)max and X(w)min corresponded respectively to water and a highly turbid honey sample being screen-illuminated in the cuvette.
For polarised light, X(w)max and X(w)min corresponded respectively to a highly concentrated sugar solution and water in the cuvette being addressed by the screen illumination via two polarising filters (Figure 5.2) aligned at 45° to each other.
For fluorescence, X(w)max and X(w)min corresponded to a high-purity honey sample and water in the cuvette being addressed by the LED light.
Chromatic Interpretation of Test Results
The normalised test data (transmission, polarisation, fluorescence) for 21 different honey samples from various sources and various degrees of purity, quality and so on were processed to display chromatically for qualitative visual comparison as well as on chromatic maps.
Chromatic cluster maps of the processed honey results were formed as graphs of R:B and G:B. Such maps were produced for the transmission (Rt:Bt), polarisation (Rp:Bp) and fluorescence (Gf:Bf) data (Figure 5.4a-c). Based upon calibration results with mixtures of honey, water and syrup, sector boundaries were produced on these maps to classify honey samples from different sources and conditions. The sectors on the transmission map (T1-T5) distinguished clear from turbid honeys; on the polarisation map (P-P3) low, high-sugar and sugar-diluted honeys, whilst the Fluorescence map (F1-F2) distinguished pure (FI) and impure (F2) honeys. Clustering of various honey samples on the three chromatic maps (transmission, polarisation, fluorescence) was interpreted as follows.
Transmission R:B Map
Region T1 honey samples were optically transparent due to a high liquid content, indicating water dilution during processing and exposure to overheating or overfifiering, that is, processed/adulterated honeys. See Figure 5.4a.
Region T2, T3 honey samples were moderately clear (T2) or had low turbidity (T3), and as such were raw honeys with low or moderate pollen contents, depending upon the floral source (Al-Zoreky et al., 2001).
Region T4 honey samples were highly turbid with lower light transmission, which suggested a long shelf life and high storage temperature (White, 1975; Gonzales et al., 1999), leading to chemical instability, botanical source or crystallisation.
Region T5 honey samples showed little optical transmission or reflection, possibly due to their mineral content (Gonzalez-Miret et al., 2005).
Polarisation R:B Map
Region PI honey samples had relatively high levels of polarised short- and long-wavelength light caused by high sugar levels (glucose/fructose) with w'ater content. See (Figure 5.4b).
FIGURE 5.4 Primary chromatic maps for various honey samples (-------classification boundaries) (a) Transmitted
light parameter Rt:Bt. (b) Polarised light parameter Rp:Bp. (c) Fluorescence parameter Gf:Bf (Sufian, 2014).
Region P2 honey samples had a higher fraction of polarised long-wavelength light due to glucose/ fructose being major constituents (Al-Zoreky et al., 2001).
Region P3 honey samples had relatively low levels of polarised light, indicating low sugar (glucose/fructose) content.
Fluorescence G:B Map
Region FI honey samples (which were of moderate purity) had fluorescence spectra shifted towards the medium wavelengths (G) due to higher ash content (Al-Zoreky et ah, 2001) and are traditionally regarded as genuine/good-quality honey (i.e.. Grade 1). See (Figure 5.4c).
Region F2 honey samples had high levels of impurities and high water content. F2 samples, which also lie in Regions T4 and T5 of Figure 5.4a, had been overheated or had invert sugar (cane sugar).
Quantification of Performance
The chromatic signatures of a honey sample on each of the three chromatic maps (Figure 5.4a-c) were combined to provide an overall empirical quantification of the honey quality by allocating a quality score to each of the different map sectors. Each sector, T1-T5, on the transmission map (Figure 5.4a) carried a score 0-1, with 0 for poor quality and 1 for pure honey. Likewise, sectors P1-P3 on the polarisation map (Figure 5.4b) were scored 0-1, and sectors F1-F2 on the fluorescence map (Figure 5.4c) were scored 0-2. The three scores for a honey sample (T, P, F) from each map were added to give an overall quality indication (T + P + F), with 0 being poor and 4 being very good.
The overall score for each honey sample was compared with the quoted formal grade (Poor = 1, Moderate = 2, Good = 3) of the honey provided by the honey producer. The honey quality predicted chromatically was compared statistically w'ith the formal grading in terms of sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) (Appendix 5A). The analysis showed that high-quality honey samples could be identified with a sensitivity of 91%, a specificity of 80%, a positive predictive value of 83% and a negative predictive value of 89%, and that poor-quality honey samples were identified with a sensitivity of 75%, a specificity of 92%, a positive predictive value of 86% and a negative predictive value of 86% (Sufian, 2014).
Overview and Summary
Chromatic techniques have been used for monitoring honey quality and adulteration using a personal computer, its visual display unit as one illumination source and a connected webcam for capturing the chromatic signatures of a honey sample. The system is flexible in being operable in the three optical domains of light, transmission/absorption, polarisation and fluorescence.
Chromatic maps for each of these three domains were produced and calibrated with mixtures of water, syrup and honey before being used for discriminating between various honey samples.
Calibration, normalisation and ambient light compensation procedures were developed to allow operation under a range of illumination conditions such as in the field (sunlight) and at commercial premises.
The approach was shown to perform well via a statistical comparison of results with formal honey classification.
The self-contained, flexible, cost-effective nature of the system and its portability enabled preliminary tests to be undertaken in remote rural areas of Yemen (Sufian, 2014), w'here several of the honey samples were tested (Figure 5.5).
FIGURE 5.5 PC-based honey monitoring system used onsite in Yemen, (a) Geographical map of Yemen with the field-test sites; (b) onsite honey monitoring with a PC in Tubasha’a Village. Sabir Mount. Taiz governorate - Midlands, Yemen; (c) onsite honey monitoring with a portable PC in a local honey shop, Sanaa, North - Yemen (Sufian. 2014).
The supply of honey samples by Dr. Eida Alssadi from the Center of Food and Medicine in Sana’a, Yemen, is acknowledged. Support provided by the British-Yemeni Society for field tests is also appreciated.
Appendix 5A: FORMULAE FOR SENSITIVITY, SPECIFICITY, POSITIVE PREDICTED VALUE, NEGATIVE PREDICTED VALUE
Sensitivity (Sn), specificity (Sp), positive predicted value (PPV) and negative predicted value (NPV) are defined by the following equations (e.g., Deakin et al., 2014).
where TP and FP are respectively the number of true and false positive chromatic results, and TN and FN are respectively the number of true and false negative chromatic results.
Al-Zoreky, N.. Alza’aemy, A. and Alhumiari, A. 2001. Quality spectrum of Yemeni honey. Journal of Agricultural Science, Damascus University, 17(2), p. 110.
Anklam. E. 1998. A review of the analytical methods to determine the geographical and botanical origin of honey. Food Chem, 63(4), pp. 549-562.
Bogdanov, S. et al. 1999. Honey quality and international regulatory standards: Review by the international honey commission. Bee World. 80(2), pp. 61-69.
Deakin, A.G., Jones. G.R., Spencer, J.W.. Bongard, E.J.. Gal, M., Sufian. A.T. and Butler. C.C. 2014. A portable system for identifying urinary tract infection in primary care using a PC-based chromatic technique Physiol. Meas. 35(5), pp. 793-805.
El-bialee. N.. Rania. К.1.. El-bialee, A.. Harith, M.A. and Darwish, N.E.M. 2013. Discrimination of honey adulteration using laser technique. Australian Journal of Basic and Applied Sciences, 7(11), pp. 132-138. Gonzales, A.P.. Burin. L. and Buera. M.D.P. 1999. Color changes during storage of honeys in relation to their composition and initial color. Food Res Int, 32(3), pp. 185-191.
Gonzalez-miret, M.I., Heredia. F.J.. Terrab. A.. Hernanz, D. and Fernandez-recamales, M.A. 2005. Multivariate correlation between color and mineral composition of honeys and by their botanical origin.J Agric Food Chem, 53(7), pp. 2574-2580.
Isengrad. H.. Schultheiss, D.. Radovic, B. and Anklam. E. 2001. Alternatives to official analytical methods used for the water determination in honey. Food Control, 12(7), p. 459.
Martin, P. and Bogdanov, S. 2002. Honey Authenticity: A Review. Swiss Bee Research Centre. Dairy Research Station, Liebefeld: Q. P. Services, Hayes, Great Britain.
Nanda. V.. Sarkar, B.C.. Sharma, H.K. and Bawa, A.S. 2003. Physico-chemical properties and estimation of mineral content in honey produced from different plants in Northern India. J Food Compos Anal, 16(5), pp. 613-619.
Ozbalci. B.. Boyaci, I.H.. Topcu, A., Kadilar, C. and Tamer, U. 2013. Rapid analysis of sugars in honey by processing Raman spectrum using chemometric methods and artificial neural networks. Food Chem, 136(3-4), pp. 1444-1452. 
Pilizota, V. and Tiban. N. 2009. Advances in Honey Adulteration Detection. [Food Safety Magazine], [online]. Available: http://www.foodsafetymagazine.com/magazine-archivel/augustseptember-2009/ advances-in-honey-adulteration-detection/
Roshan. A. et al. 2013. Authentication of monofloral Yemeni sidr honey using ultraviolet spectroscopy and chemometric analysis. J Agric Food Client, 61(32). pp. 7722-7729.
Sufian. A.T. 2014. Honey monitoring in the Yemen. British-Yemeni Society Journal. 22, pp. 50-57.
Terrab. A.. Diez. M.J. and Heredia. F.J. 2002. Chromatic characterisation of Moroccan honeys by diffuse reflectance and tristimulus colorimetery- non-uniform and uniform colour spaces. Journal of Food Science and Technology International, 8(4), pp. 189-195.
White, J.W. 1975. Composition of Honey. Honey: A Comprehensive Survey. Heinemann: London, UK. pp. 157-206.