Damage Detection and Classification of Mangoes Based on Thermal Imaging

Abstract

Thermal imaging is an effective tool to measure the surface temperatures of any object. Nowadays, it is widely used to evaluate the quality of different agricultural products. This work uses thermal images to assess mechanical damage in mangoes. During the image acquisition, hot air was supplied to ensure the accuracy of the image. However, the quality of the images was not sufficient to distinguish the subtle and noticeable variations in damage for the next few days. Therefore, each mango image's grey level co-occurrence matrix for consecutive days was examined, and the deviation was plotted. A total of 4505 samples comprised 3226 training samples, and 1279 testing samples obtained from 400 mangoes over ten days were processed with a modified deep learning method. As a result, a 98.3% accuracy rate was achieved.