Image-based date fruit classification

Date fruits are small fruits that are abundant and popular in the Middle East, and have growing international presence. There are many different types of dates, each with different features. Sorting of dates is a key process in the date industry, and can be a tedious job. In this paper, we present a method for automatic classification of date fruits based on computer vision and pattern recognition. The method was implemented, and empirically tested on an image data spanning seven different categories of dates. In our method, an appropriately crafted mixture of fifteen different visual features was extracted, and then, multiple methods of classification were tried out, until satisfactory performance was achieved. Top accuracies ranged between 89% and 99%.