ReseaRch PaPeR Journal of Agricultural and Marine Sciences Vol. 22 (1): 36-41 DOI: 10.24200/jams.vol22iss1pp36-41 Reveived 10 Jun 2016 Accepted 15 Nov 2016 Computer vision technique to classify dates based on hardness Naeema H. Al-Shekaili1, Annamalai Manickavasagan2*, Nawal K. Al-Mezeini2, M. Shafiur. Rahman1 and Negib. Guizani1 2*Annamalai Manickavasagan ( ) Sultan Qaboos University, College of Agricultural and Marine Sciences, Department of Soils, Water and Agricultural Engineering. Box 34, Al-Khod 123. Sultanate of Oman. current email: mannamal@uoguelph.ca. 1 Sultan Qaboos University, College of Agricultural and Marine Scienc- es, Department of Food Science and Nutrition. Introduction Date palm (Phoenix dactylifera, L.) is one of the oldest and most important staple crops in the Middle East and North Africa. The Sultanate of Oman is ranked among the top ten date producing coun- tries in the world with approximately 49% of cultivable land area (FAO 2010). Although date production was 276,400 MT in 2010 (FAO, 2010), only about 5,000 tons was exported (Zaabanot, 2011). The low level of export from Oman is due to improper sorting of dates to ensure higher quality as expected by consumers (Al-Marshudi, 2002). According to the CODEX standard, the quality attributes to grade dates are colour, flavor (sugar level), moisture content and absence of defects, such as insect damage, and surface damage (Kader and Hussein, 2009). Moisture content in dates is important because ex- cessive loss of water causes drying, consequently they becomes hard (Kader and Hussein, 2009; Rahman and Al-Farsi, 2005). Hardness beyond a critical value is con- sidered as a defect in dates as it affects the physical prop- erties and consumer acceptability. Dates can be subdi- vided into: soft, semi-dry (semi-hard) and dry (hard) according to their moisture content or hardness (Kad- er and Hussein, 2009; Al-Janobi, 1998). Hard dates are chewy and tough with strong curvy and zigzag textured skin (Rahman and Al-Farsi, 2005). Dates are processed into different products and the choice of the date types for a given product depends on the final product. For example, soft dates are used to manufacture date syr- استخدام تقنية تصوير احلاسوب يف تصنيف التمور حسب الصالبة نعيمة حارب الشكيلي1 ومنيكافا سجان أنامايل2* ونوال مخيس املزيين2 وحممد شفري رمحان1 وجنيب قيزاين1 Abstract. Hardness is one of the important attributes in determining the quality of dried fruits. Hardness assessment is normally carried out by manual inspection. This method is time consuming, laborious, expensive and subjective. The objective of this study was to develop a computer vision system with a monochrome camera to classify dates based on hardness. Date samples were obtained from three different growing regions in Oman and graded into soft, semi-hard, and hard classes based on hardness. A total of 1800 date samples were imaged individually using a monochrome camera (600 dates / class). Histogram and texture features were extracted from the acquired monochrome images and used in the classification models. The overall classification accuracies in three class model (soft, semi-hard, and hard) were 66% and 71% for linear discriminant analysis (LDA) and artificial neural network (ANN), respectively. It was improved to 84% and 77% in LDA and ANN, respectively while using two class model (soft and hard (semi-hard and hard together)). The histogram features were more contributing in the date classification based on hardness than image texture features. Computer vision technique has great potential to develop online quality monitoring systems for dates and other dried fruits. Keywords: Dates; hardness; histogram features; texture features; gray scale images امللخــص: تعتــر الصالبــة مــن أهــم اخلصائــص يف تقييــم جــودة الفواكــة اجلافــة. ويتــم تقييــم الصالبــة عــادة عــن طريــق التفتيــش اليــدوي، إال أن هــذه الطريقــة تتطلــب اجلهــد والوقــت الطويــل، كمــا أهنــا باهضــة وغــري موضوعيــة. هتــدف هــذه الدراســة إىل تطويــر أســلوب التصويــر باســتعمال احلاســوب متصــال بكامــريا تصويــر أحاديــة اللــون هبــدف تصنيــف التمــور اعتمــاداً علــى الصالبــة. وقــد مت احلصــول علــى عينــات التمــور مــن 3 مناطــق خمتلفــة بســلطنة عمــان وصنفــت هــذه التمــور إىل 3 فئــات: لينــة، شــبه صلبــه، وصلبــه. حيــث مت تصويــر جممــوع 1800 عينــة )600 عينــة لــكل فئــة( باســتعمال كامــريا التصويــر أحاديــة اللــون. وقــد مت اســتخراج مالمــح قــوام التمــور مــن الرســم البيــاين واســتخدمت كنمــوذج تصنيــف. وقــد حققــت إمجــال تصنيــف التمــور باســتعمال منــوذج الثــالث فئــات )اللينــة، شــبه صلبــه، وصلبــه( دقــة تقــدر بـــ66% و 79% عنــد اســتعمال طريقــة التحليــل اخلطــي التمييــزي )LDA( و طريقــة التحليــل اخلطــي املتــدرج )ANN( علــى التــوايل. كمــا حققــت النتائــج نســب دقــة أعلــى عنــد اســتعمال منــوذج الفئتــن )فئــة لينــة، وفئــة شــبه صلبــه، وصلبــه معــا( تقــدر بـــ 84% و 77% بطريقــة )LDA( و)ANN( علــى التــوايل. وقــد كانــت مســامهة مالمــح الرســم البيــاين أوضــح مــن مالمــح القــوام يف تصنيــف التمــور. ميتلــك التصويــر باســتعمال احلاســوب قــدرات عاليــة ميكــن اســتخدامها لتطويــر أنظمــة مراقبــة جــودة التمــور والفواكــة اجلافــة األخــرى عــر األنرتنــت. الكلمات املفتاحية: التمور ، الصالبة ، مالمح الرسم البياين ، مالمح القوام ، صور املقياس الرمادي 37Research Article Al-Shekaili, Manickavasagan, Al-Mezeini, Rahman, Guizani up while hard dates are used to produce flour. There- fore hardness is one of the important parameters used to evaluate and classify dates in industry. The presence of hard dates in other grades affects the acceptability of the whole batch and yields low values in domestic and international markets. In general, hardness assessment is carried out by the traditional method of visual inspection or mechanical methods like diagonal metal plate method or vacuum systems (Huxsoll and Reznik, 1969; Chesson et al., 1979; Al-Janobi, 2000). The visual inspection method is sub- jective, laborious and expensive, whereas mechanical methods are sample-destructive nature and conducted only on representative samples. An objective non-de- structive method for sorting of dates based on hardness would be highly beneficial for online quality assessment and monitoring of dates in handling facilities. Therefore the objective of this study was to determine the ability of a computer vision system with a monochrome camera to classify dates based on hardness. The monochrome camera was selected because of its low cost, image size and faster image handling and processing capabilities. Computer based image processing techniques or computer vision technologies replace the traditional method of human inspection towards achieving bet- ter, faster, and automated operations (Patel et al., 2012; Pour-Damanab et al., 2012). It gives a meaningful de- scription for the object by duplicating human vision us- ing different algorithms to assess the quality (Narendra and Hareesh, 2010). It is used to characterize colour, tex- ture and complex geometric properties (Chandraratne et al., 2003). Computer vision is used in food industry for the classification, quality evaluation, sorting, grading, and defect detection (Du and Sun, 2006; Brosnan and Sun, 2004). The application of computer vision in the date industry is scarce (Lee et al., 2008a). Schmilovitch et al. (1999) developed a semi-automated vision system for maturity determination of fresh dates using NIR scanner. Al-Janobi (2000) graded Saudi dates (Sifri va- riety) based on colour and texture with an average error of 1.8% using a colour camera. A computer vision sys- tem for the grading of dates based on fruit size and skin delamination using reflected NIR imaging showed 10% higher accuracy over human inspection and a reduction in labor cost by 75% (Lee et al., 2008b). In another study, using a RGB colour imaging system to grade dates into 3 categories based on size, shape, flabbiness intensity and defects yielded 80% accuracy (Al-Ohali, 2011). Materials and methods Sample collection Samples of dates of the Fard variety, the most processed variety in Oman, were obtained from three major dates growing regions: Al-Batinah, Al-Dakhliah and Al-Shar- qiah. The dates were sorted into three classes based on hardness (hard, semi-hard and soft) by an experienced grader and confirmed by a manager in Bright Sun Dates Company, Oman. From each region, a representative sample of 600 dates (200 per class), in total 1800 sam- ples, were selected and used in this study. Image acquisition system A monochrome camera (model: XCD-X700, Sony, Ja- pan) was used to acquire uncompressed 8-bit images of resolution 1024×768 pixels with a charge coupled device (CCD) sensor. It was placed at a height of 1 m in order to simulate the application in handling facilities. Images were taken in a dark room while illuminating the sample with two fluorescent lights (36 W, model: Dulux L, OS- RAM, Italy). The camera was connected to the computer through IEEE 1394 cable. The camera was calibrated us- ing white and black standard colour cards (Digital Kard XL, DGK colour Tools, USA) before starting each batch of imaging. Image analysis and classification model The region occupied by date sample in the monochrome image was segmented from the background using Matlab (version 7.6.0) (Fig. 1). Histogram and texture features of the grayscale information for the region pertaining to date sample were extracted and analyzed statistically (at α ≤ 0.05). The details of the extracted features are given in Table 1. The classification accuracy of a linear discriminant analysis (LDA) and stepwise linear discriminant anal- ysis (SLDA) was performed with SPSS software. Simi- larly the accuracy of a back propagation neural network (BPNN) was obtained using Matlab. Discussion Although date samples used in this study belonged to the same variety, regional differences existed in their Figure 1. Process involved in the segmentation of dates from the background from left to right (a) original monochrome image (b) adjusted monochrome image (c) binary image (d) binary image after the morphological operations to reduce noise (e) segmented date. 38 SQU Journal of Agricultural and Marine Sciences, 2017, Volume 22, Issue 1 Computer vision technique to classify dates based on hardness external properties such as colour, shape and size. In general, hard dates were brighter in colour compared to soft and semi-hard dates. However, there were several overlaps in gray scale values across different classes and regions and interfered in the classification. Features extracted from monochrome images Histogram features The histogram features of the date samples from differ- ent regions are shown in Table 2. There was a difference between soft, semi-hard and hard dates in the mean gray value. The maximum and minimum mean gray values were associated with hard and soft dates, respectively. This indicated that the hard dates were brighter in co- lour compared to the soft and semi hard dates. There was no difference in the mean gray value of the dates from the Al-Dakhliah and Al-Batinah regions. Standard deviation and variance of the soft, semi-hard and hard dates varied significantly. It was the highest for soft dates and the lowest for hard dates. There was no difference in the standard deviation and variance of the dates from Al-Batinah and Al-Sharqiah regions. The smoothness was different between the three classes of dates with maximum and minimum values for soft and hard dates, respectively. However, there was no difference in smoothness between dates from the Al-Sharqiah and Al-Dakhliah regions. There was a difference in eccentricity between three classes with hard dates having the highest values and soft dates the lowest values. The eccentricity of the dates was not dif- ferent between the Al-Batinah and Al-Sharqiah regions. The growing regions produced differences in solidity of date samples. The extent of three classes of dates was significantly different. The highest and lowest value was obtained for hard and soft dates, respectively. Table 1. Features extracted from monochrome images. Features Definition* Histogram features Mean gray value Average of the gray values of all the pixels in an image Standard deviation Standard deviation of all the pixels in an image Variance Variance of all the pixels in an image Smoothness Measure of the relative smoothness of the intensity in a region Eccentricity Ratio of distance between the foci of the ellipse and its major axis length Solidity Proportion of the pixels in the con- vex hull that are in a region Extent Proportion of the pixels in the bounding box that is in the region Texture features (GLCM**) Contrast Measure of contrast between a pixel and its neighbor over the whole image Correlation Measure of how correlated a pixel is to its neighbor over the whole image Energy Sum of squared elements in the GLCM Homogeneity Similarities of pixels Maximum probability Maximum occurrence of the gray level Entropy Measure of the randomness of intensity image Cluster prominence Measure of the skewness of a matrix Cluster shade Measure of the lake of symmetry Dissimilarity Measure of the dissimilarity be- tween the pixels * Manickavasagan et al. (2008a,b); Basavaraj and Vishwanath (2009); Gonzalez et al. (2010) ** Gray Level Co-occurrence Matrix Table 2. Mean values of histogram features extracted from monochrome images of dates (n=200). Feature Region Al-Batinah Al-Dakhliah Al-Sharqiah Soft Semi- hard Hard Soft Semi- hard Hard Soft Semi- hard Hard Mean gray value 45.69 48.34 57.16 45.88 49.03 56.38 47.50 53.84 61.91 Standard deviation 39.26 37.34 33.44 36.07 35.08 33.98 39.62 36.22 32.90 Smoothness 0.9993 0.9992 0.9990 0.9992 0.9991 0.9990 0.9993 0.9992 0.9989 Eccentricity 0.7754 0.7820 0.7958 0.7950 0.8127 0.8089 0.7633 0.7861 0.7959 Solidity 0.9867 0.9871 0.9866 0.9862 0.9861 0.9864 0.9848 0.9861 0.9861 Extent 0.7996 0.8047 0.8124 0.8096 0.8134 0.8164 0.7959 0.8048 0.8160 39Research Article Al-Shekaili, Manickavasagan, Al-Mezeini, Rahman, Guizani Texture features The values of texture features for dates are given in Ta- ble  3. The soft dates had more contrast than the semi- hard and hard dates. However, there was no difference in the contrast of semi-hard and hard dates. It was max- imum for Al-Batinah and minimum for Al-Sharqiah re- gions. The correlation of dates varied significantly with respect to classes and regions. There was no difference in the energy between the three classes. However, there was a difference in the energy between the three regions. The homogeneity of date samples from three regions was different with maximum and minimum values for Al-Sharqiah and Al-Batinah regions, respectively. How- ever, the homogeneity of hard and semi-hard dates was not different. There was no difference in the maximum probability between three classes of dates. On the other hand, the maximum probability between the three re- gions was significantly different. The Al-Sharqiah region had the highest and Al-Dakhliah had the lowest proba- bility. The entropy, cluster prominence and cluster shade of the date samples from three regions were different. The Al-Dakhliah region had the highest entropy, clus- ter prominence and cluster shaded while Al-Sharqiah had the lowest values. However, there was no difference in the entropy, cluster prominence and cluster shaded between the three classes of dates. The soft dates were more dissimilar in comparison to semi hard and hard dates. But there was no difference in the dissimilarity of semi-hard and hard dates. Table 3. Mean values of texture features extracted from monochrome images (n=200). Feature Region Al-Batinah Al-Dakhliah Al-Sharqiah Soft Semi- hard Hard Soft Semi- hard Hard Soft Semi- hard Hard Contrast 0.0246 0.0241 0.0232 0.0235 0.0230 0.0240 0.0222 0.0221 0.0215 Correlation 0.9956 0.9956 0.9956 0.9956 0.9957 0.9958 0.9951 0.9953 0.9953 Energy 0.8839 0.8859 0.8893 0.8876 0.8876 0.8789 0.9060 0.9017 0.9038 Homogeneity 0.9995 0.9995 0.9995 0.9995 0.9995 0.9995 0.9996 0.9996 0.9996 Maximum probability 0.9381 0.9392 0.9411 0.9402 0.9402 0.9352 0.9505 0.9482 0.9493 Entropy 0.2349 0.2317 0.2262 0.2291 0.2291 0.2424 0.1992 0.2063 0.2027 Cluster prominence 1824.32 1798.46 1754.3 1779.45 1779.99 1883.85 1532.95 1592.93 1562.43 Cluster shade 138.34 136.25 132.71 134.66 134.70 143.259 114.96 119.70 117.34 Dissimilarity 0.00351 0.00344 0.00331 0.003367 0.0032 0.00343 0.00318 0.00316 0.00307 Table 4. Classification accuracy (%) of date samples in three and two class models. Region LDA* SLDA** Selected features for SLDA Three-class model Al-Batinah 59 58 mean gray value, variance and extent Al-Dakhliah 67 66 mean gray value, eccentricity, smoothness, cluster prominence and maximum probability Al-Sharqiah 76 75 mean gray value, variance, extent, correlation and smoothness All regions together 66 66 mean gray value, standard deviation, variance, extent, correlation, smoothness, dissimilarity and maximum probability Two-class model Al-Batinah 83 83 mean gray value, variance and extent Al-Dakhliah 87 87 mean gray value, variance, solidity, extent , smoothness, cluster prominence and maximum probability Al-Sharqiah 86 85 mean gray value, standard deviation and extent All regions together 83 84 mean gray value, standard deviation, solidity, extent, smoothness, entropy, maximum probability * all sixteen features were used in the linear discriminant analysis (LDA) ** most contributing features were used in the stepwise linear discriminant analysis (SLDA) 40 SQU Journal of Agricultural and Marine Sciences, 2017, Volume 22, Issue 1 Computer vision technique to classify dates based on hardness Classification models Linear discriminant analysis (LDA) The features extracted from the monochrome images were used to determine the efficiency of this technique in sorting of dates based on hardness. In the first approach, the date samples were classified into three groups name- ly soft, semi-hard and hard (three class model). In some applications, dates are graded into only two categories such as soft and hard. Therefore in the second approach, the date samples were classified into two groups namely soft and hard (semi hard and hard together) (two class model). Analyses were also performed by considering each region separately and combined together. Table 4 shows the accuracies obtained in different approaches. In three class models, it was in the range of 58% to 76%. The highest and lowest classification accuracies were obtained for the Al-Sharqiah and the Al-Batinah regions, respectively. While analyzing all regions together, there was no difference between LDA and stepwise linear discriminant analysis (SLDA) with most contributing factors. The SLDA selected mostly histogram features for the classification. This indicates that histogram features are more contributing than the texture features in the classification of dates based on hardness. Similarly, Basavaraj and Vishwanath (2009) reported that texture features including contrast, cor- relation, energy, entropy, homogeneity and dissimilarity were not sufficient for the classification of bulk sugary foods. On the other hand, Chandraratne (2003) reported that image texture features were suitable indicators for beef tenderness because the R2 value was 0.621 while using geometrical features and 0.746 with texture fea- tures. Also, Li et al. (1999) and Li et al. (2001) mentioned the same about the importance of texture features in classification of beef tenderness. In two class models, the classification accuracy was 83% to 87%. The highest classification in this approach was achieved for Al-Sharqiah region. Zayas et al. (1996) obtained 63% accuracy for hard wheat and 91% accuracy for soft wheat in two class mod- el, using monochrome images. Li et al. (2001) classified the steaks using colour camera into rough and tender with an accuracy of 83%. Artificial neural network (ANN) The classification accuracies of ANN for three class model are shown in Table 5. The misclassification was observed between the soft and semi-hard and hard and semi hard dates. The overall accuracy obtained was 71%. Fadel (2007) used colour camera and obtained a classi- fication accuracy of 100%, 80%, 80%, 60% and 80% for Fard, Khalas, Lolo, Bomaan and Berhi dates varieties, respectively using probabilistic neural network. While grading the dates into three grades according to size, shape, flabbiness intensity, and defects using RGB im- ages, Al-Ohali (2011) obtained 55% to 80% classification accuracy. In two class model, the classification accuracy of the hard dates was higher than soft dates (Table 6). The overall accuracy in the two class model was 77%. Conclusion A computer vision system with monochrome camera was developed to classify dates based on hardness with varying degree of success. The classification was car- ried out with histogram and texture features extracted from the monochrome images of dates using LDA and ANN. The classification accuracies of two class models were higher than three class models in both LDA and ANN. In SLDA, histogram features contributed more for classification than texture features. The potential of computer vision technique for hardness determination in dates must be investigated with other cameras such as NIR and RGB colour cameras to improve classification accuracy. Acknowledgements The authors would like to acknowledge The Research Council (TRC) in Sultanate of Oman for funding this work under the project “Development of computer vi- sion technology (CV) for the quality assessment of dates in Oman”. References Al-Janobi, A. 1998. Application of co-occurrence matrix method in grading date fruits. Paper No. 98 – 3024. ASAE meeting presentation, King Saud University. Al-Janobi, A. 2000. Date inspection by color machine vi- sion. Journal of King Saud University of Agricultural Science 12: 69 – 79. Al-Marshudi, A.S. 2002. Oman traditional date palms: Production and improvement of date palms in Oman. Tropicultura 20: 203 – 209. Al-Ohali, Y. 2011. Computer vision based date fruit Table 5. Accuracies (%) of ANN to classify dates according to hardness in the 3-class model. From / To Soft Semi-hard Hard Soft 76 22 2 Semi-hard 17 68 15 Hard 5 25 70 Table 6. Accuracies (%) of ANN to classify dates according to hardness in the 2-class model. From / To Soft Hard Soft 69 31 Hard 16 84 41Research Article Al-Shekaili, Manickavasagan, Al-Mezeini, Rahman, Guizani grading system: Design and implementation. Journal of King Saud University, Computer and Information Sciences 23: 29 - 36. Basavaraj, S.A. and Vishwanath, C. B. 2009. Texture based identification and classification of bulk sugary food objects. ICGST-GVIP J 9: 9-14. Brosnan, T. and Sun, D. 2004. Improving quality inspec- tion of food products by computer vision-a review. Journal of Food Engineering 61: 3-16. Chandraratne, M.R., Samarasinghe, S., Kulasiri, D., Frampton, C., Bekhit, A.E.D. and Bickerstaffe, R. 2003. Prediction of lamb tenderness using texture features. Research report, Lincoln University, ISSN 1174 - 6696. Chesson, J.H., Burkner, P.F. and Perkins, R.M. 1979. An experimental vacuum separator for dates. Transac- tions of ASAE, 22: 16 - 20. Du, C. and Sun, D. 2006. Learning techniques used in computer vision for food quality evaluation: a review. Journal of Food Engineering 72: 93 - 55. Fadel, M. 2007. Date fruits classification using proba- bilities neural networks. Agricultural Engineering International: The CIGR Journal (CIGR Commission Internationale du Genie Rural J). Manuscript ID 07 003. IX: 1-10. FAO. 2010. http://faostat.fao.org/site/339/default.aspx. Access done on 14th January 2013. Gonzalez, R.C., Wood, R.E. and Eddins, S.L. 2011. Dig- ital image processing using MATLAB. New Delhi: Tata McGraw Hill Education Private Limited. Huxsoll, C.C. and Reznik, D. 1969. Sorting and process- ing mechanically harvested dates. Date Grower’s In- stitute Report 46: 8 - 10. Kader, A. and Hussein, M. 2009. Harvest and posthar- vest handling of dates. Project on the development of sustainable dates palm production system in the GCC countries of Arabian Peninsula. International Center for Agricultural Research in the Dry Areas. Lee, D., Archibald, J. , Chang, Y. and Gerco, R. 2008a. Robust color space conversion and color distribution analysis technique for date maturity evaluation. Jour- nal of Food Engineering 88: 364 - 372. Lee, D., Schoenberger, R., Archibald, J. and McCollum, S. 2008b. Development of a machine system for auto- matic date grading using digital reflective near-infra- red imaging. Journal of Food Engineering: 388 - 398. Li, J., Tan,J., Martz, F.A. and Heymann, H. 1999. Image texture feature as indicators of beef tenderness. Meat Science 53, 17 - 22. Li, J., J. Tan and P. Shatadal. 2001. Classification of tough and tender beef by image texture analysis. Meat Sci- ence 57: 341 - 346. Manickavasagan, A., Sathya, G. and Jayas,D.S. 2008a. Comparison of illumination to identify wheat classes using monochrome images. Computer and Electron- ics in Agriculture 63: 237-244. Manickavasagan, A., Sathya, G., Jayas, D.S. and White, N.D.G. 2008b. Wheat class identification using monochrome images. Journal of Cereal Science 47: 518-527. Narendra, V.G. and Hareesh, K.S. 2010. Quality inspec- tion and grading of agricultural and food products by computer vision – a review. International Journal of Computer Application 2: 43-65. Patel, K.K., Kar, A., Jha, S.N. and Khan, M.A. 2012. Ma- chine vision system: a tool for quality inspection of food and agricultural products. Journal of Food Sci- ence and Technology 49: 123-141. Pour-Damanab, A.S., Jafary, A. and Rafiee, S. 2012. Ki- netics of the crust thickness development of bread during baking. Journal of Food Science and Technol- ogy, doi: 10.1007/s13197-012-0872-z. Rahman, M.S. and Al-Farsi, S. 2005. Instrumental tex- ture profile analysis (TPA) of date flesh as function of moisture content. Journal of Food Engineering 66: 505 - 511. Schmilovitch, Z., Hoffman, A., Egozi, H., Ben-Zvi, R., Bernstin ,Z. and Alchanatis,V. 1999. Maturity deter- mination of fresh dates by near infrared spectrome- try. Journal of the Science of Food and Agriculture 79: 86 - 90. Zaabanot, A. 2011. Improve date syrup production in Oman – Case study. Published in Oman Daily, Satur- day 8th October, Edition No11085: 22. Zayas, I.Y., Martin, C.R., Steele, J.L. and Katsevich, A. 1996. Wheat classification using image analysis and crush-force parameters. Transactions of the ASAE, 39: 2199 - 2204.