Automatic Inspection Of The External Quality Of The Date Fruit

Abstract

Assessing the quality of date fruits manually is a labor-intensive task. Furthermore, the quality of the date fruits in storage may degrade with time and, therefore, it is important to inspect the quality of date fruits routinely. After date harvesting, industrial companies initiate the inspection of dates, a process through they isolate the damaged or defected dates from the good and healthy ones. Industrial factories demand a high-quality and fast production that can be achieved through automatic inspection.

In this study, we developed a method of inspecting the external quality of khalas date fruit through image processing. Images of date fruits were classified into good-quality fruits and sugar-defect fruits by using a Bag-of-Feature (BOF). The methodology comprises five main stages including key-points detection, feature(s) extraction, creating a dictionary, vector quantization, and classification. The developed framework was tested on the dataset that was collected by the authors using two types of key-points detection methods. The best classification accuracy for the best grid search was 99% and 84% for the SURF-based key-points detection method on the testing dataset. This research we published in ELSEVIER.

Keywords:

computer vision, fruit classification, bag of feature, machine learning, date inspection, k mean clustering, SURF descriptor, SVM supported vector machine, error correcting output codes (ECOC), date fruit