Disease detection and physical disorder classification in citrus fruit

Citrus fruit production takes place in 133 countries around the world and Turkey is one of the countries with a critical position in respect of the production. According to TSI 2019 statistics, all 4.3 million tons of citrus production in Turkey has taken place in the Mediterranean and Aegean regions. Correlatively to TSI (Turkey Statistical Institute) 2020 statistics, Turkey ranks 7th with 1.9 tons of orange production. The detection of production and quality loss gained importance, out of consideration of these high production rates. About 50% of citrus fruits produced are wasted due to various diseases and physical defection every year. When the disease analysis is not accomplished, it has effects ranging from environmental pollution to tree loss and economic losses as a result. Therefore, to prevent such threats, different artificial intelligence techniques started to be used in this field. In this study, physical disorders of citrus fruit are classified into two classes named defected and deformed for the classification problems. The defected class contains 2379 data of environmentally damaged oranges (e.g., scab, frost damage). The other class called deformed contains 2770 data of physically damaged oranges (damage caused by transport or branch wounds). Similar to the classification problem, the object detection problem contains two classes of diseases. Frequently encountered disease called Alternaria alternata sp. Citri and a frequent pest disease called Thrips are chosen for the problem. The number of 1681 images are collected for Alternaria alternata sp. Citri and 1901 images for Thrips. Disease recognition has been carried out with the proper method.