RGB color imaging to detect Aspergillus flavus infection in dates

Abstract

In this study, the potential of RGB color imaging to detect fungal contamination in three varieties (Khalas, Fard and Naghal) of dates was
investigated. The samples were treated as three groups: UC (untreated control), SC (surface sterilized, rinsed and air-dried) and IS (surface
sterilized, rinsed, air-dried and fungal inoculated). Color images of control samples and A. flavus inoculated date fruits after every 48 h of
inoculation for 10 days were acquired using an RGB color imaging system (n=3150). The classification accuracies for IS were compared
with UC and SC separately using a two-class model (control vs. infected (all stages of infection together)), six-class model (control,
infected day 2, day 4, day 6, day 8 and day 10) and pair-wise model (control vs. each stage of infection). In the two-class model, the
highest accuracy obtained by Fard, Khalas and Naghal dates were 97, 100 and 99%, respectively. The developed algorithm was also
tested on pooled dates (all three varieties were combined together: Control vs infected), and 98% and 99% of infested samples were
correctly classified from untreated control and sterile control, respectively in two class models. In six class models, highest classification
accuracies of 99-100% were obtained for IS Day 10 in all three date varieties