English abstract
The important quality attributes to evaluate date fruits are color, flavor, moisture content and absence of defects. Hardness is considered to be a defect in dates mainly caused by lower moisture content. It affects the appearance and consumer acceptability. The ability of monochrome imaging technique to classify dates based on their hardness was investigated. Dates are classified into three classes; soft, semi-hard, and hard, based on hardness. A total of 1800 Fardh variety (600/class) were obtained from Al-Batinah, Al-Dakhliah and Al-Sharkiah regions (600/region). Images of individual date samples were captured using an image acquisition system. After imaging, the hardness and moisture content of each date were measured using a texture profile analyzer (TPA) and oven drying method, respectively (60/class). An algorithm was developed to extract seven histogram and nine texture features from the images using Matlab software. The classification accuracy of the monochrome imaging was obtained using linear discriminant analysis. Finally, the features extracted were correlated with the instrumental measurements of hardness and moisture content and prediction models were developed. When dates were classified into soft, semi-hard and hard classes (three classes models) with sixteen features extracted, the mean classification accuracies were 59%, 67%, 76% and 66% for Al-Batinah, Al-Dakhliah, Al-Sharqiah and three regions together respectively. However, the accuracies increased to 83% for Al-Batinah, 84% for Al-Dakhliah, 86% for Al-Sharqiah and 83% for the three regions together when classifying into soft and hard (two classes models, (soft and semi-hard dates together)). The histogram features were more contributing in the date classification based on hardness than texture features. The prediction models for hardness using the sixteen features extracted had Rvalue equal to 0.860, 0.715, 0.798 and 0.570 for Al-Batinah, Al-Dakhliah, Al-Sharqiah and combined regions together respectively. While the moisture prediction models had R' value equal to 0.43 for Al-Batinah, 0,67 for Al Dakhliah, 0.53 for Al-Sharqiah and 0.34 for combined regions together. It may be possible to develop and use monochrome imaging system in handling facilities for grading of dates based on hardness. However, the effects of differences in dates characteristic growing in other regions and countries which affects the classification must be studied.
