Study the Possibility of Physical Assessment and Grading of Mazafati Dates Using Digital Image Processing and Support Vector Machines

Abstract

The traditional methods of grading dates, due to the lack of specific characteristics, have led to incorrect grading and wasting both time and money. Date grading based on the classification algorithms could reduce seller and buyer disagreement. It also allows the product to be sold at the right price. In this research, identification of some qualitative characteristics of Mazafati dates and its classification into four categories (grades 1, 2, 3, and 4) has been done according to the opinions extracted from experts Intending to define a significant link between the quality of dates, mobile image processing application conducted in in Matlab, support vector machine (SVM) was The results of linear, quadratic, cubic, and medium Gaussian SVMs were 100% accurate, meaning that the classification had been successful. The ROC curve provided a positive classification rate versus a false positive rate for selecting classification training. A grade 1 positive rate of 0.97 indicates that the current classifier allocates 0.97 of the observations correctly to the positive class (primarily rank). In order to make the final verification, the Kappa coefficient was used. All Kappa values are greater than 0.6 and have sufficient stability. Also, the highest Kappa coefficient was related to the cubic method by more than 0.8 and the lowest one was related to fine Gaussian with a value of 0.76. Due to the accuracy and precision of implementation with SVM, this method with high efficiency wascapable of grading dates.

Keywords

• digital image processing
• grading
• qualitative features of mazafati dates
• support vector machine