FBCNN-TSA: An optimal deep learning model for banana ripening stages classification

Abstract

Ripeness of the fruit is significant in agriculture since it affects the fruit’s quality and sales. Manually determining the fruit’s ripeness has various drawbacks, including the fact that it consumes time, needs a lot of work, and occasionally results in errors. One of the crucial areas of the economies of nations is the agricultural sector. However, the manual approach is still occasionally used to assess the maturity of fruit. Fruit ripeness could be automatically categorized by the advancement of computer vision and machine learning technology. The Convolutional Neural Network (CNN) is used in this work is to classify the different ripeness stages of banana fruit. The four stages of banana ripeness are unripe, mid-ripe, ripe, and overripe. Proposed method uses a fuzzy-based convolutional neural network with tunicate swarm algorithm. The proposed model outperforms cutting-edge computer vision-based algorithms in both coarse and perfectly acceptable classification of maturation phases. The experimental results using images of bananas at various stages of ripening, achieves overall accuracy of 96.9%.