Guava fruit (Psidium guajava) damage and disease detection using deep convolutional neural networks and thermal imaging

Abstract

ABSTRACT Guava is a fruit grown predominantly in the tropical and subtropical regions of the world. Its skin is thin and soft. Although postharvest handling and transportation of fruits are unavoidable, hand harvesting is preferred to avoid physical injuries. Visually identifying internal damage is a laborious and time-consuming task. In this study, 4129 thermal and digital images of guava fruits were collected. A digital image-based convolution neural network (CNN) model was developed to classify the quality of damaged and diseased fruits. The surface temperatures of all three levels of maturity-indexed and damaged fruits were measured using thermal images. During the storage period, the temperature of the immature fruits was less than half of the mature fruits. Because the architecture was trained on dehydrated and diseased guava, it accurately predicted fruit quality. The prediction accuracy of the developed CNN model was approximately 99.92%.