Detection of Mulberry Ripeness Stages Using Deep Learning Models

Seyed-Hassan Miraei Ashtiani,Shima Javanmardi,Fons J Verbeek,Mehrdad Jahanbanifard,Alex Martynenko

doi:10.1109/access.2021.3096550

Abstract

Ripeness classification is one of the most challenging tasks in the postharvest management of mulberry fruit. The risks of microbial contamination and human error in manual sorting are significant; it may result in quality degradation and wasting of processed products. Due to advanced developments in computer vision and machine learning, automated sorting became possible. This study presents the results of developing and testing a computer vision-based application using convolutional neural networks (CNNs) for the classification of mulberry fruit ripening stages. To reduce the training cost and improve the accuracy of classification, transfer learning was used to fine-tune the CNN models. The CNN models in the test include DenseNet, Inception-v3, ResNet-18, ResNet-50, and AlexNet. Transfer learning was used to fine-tune the models and improve the accuracy of classification. The AlexNet and ResNet-18 networks exhibited the best performance with 98.32% and 98.65% overall accuracy for classifying the ripeness of white and black mulberries, respectively. Moreover, the performance of the models did not change when the data sets of both genotypes were mixed. The ResNet-18 was able to classify both genotype and ripeness from 600 fruit images in 2.36 min with an overall accuracy of 98.03%, which was superior to other architectures. It indicates that the model could be used for precise classification of the ripening stages of mulberries and other horticultural products, as a part of an automated sorting system.

Highlights

Mulberry (Morus spp., Moraceae family) is one of the fruit species, widely distributed from temperate to subtropical zones of the northern hemisphere to the tropical zones of the southern hemisphere [1], [2]
The reason for splitting the data set into two subsets is that in small data sets, the additional split might lead to a smaller training set which may be exposed to overfitting [55]
The best detection accuracy can be achieved by selecting an optimal number of epochs for training the convolutional neural networks (CNNs) models [57]

Summary

Introduction

Mulberry (Morus spp., Moraceae family) is one of the fruit species, widely distributed from temperate to subtropical zones of the northern hemisphere to the tropical zones of the southern hemisphere [1], [2]. Among the 24 known Morus cultivars, white mulberry (Morus alba L.), black mulberry (Morus nigra L.), and red mulberry (Morus rubra L.) are the most cultivated species in the world [3]–[5]. They are an excellent source of many nutritive compounds such as vitamins, minerals, polysaccharides, fatty acids, and amino acids [3], [6], as well as phenolic compounds including carotenoids, anthocyanins, flavonoids, and phenolic acids [7], [8] with the health-promoting and pharmacological effects, such as anti-cancer, anti-cholesterol, antiinflammation, anti-diabetic, anti-aging, antioxidant, antiobesity and neuroprotection [9], [10]. Mulberry is a non-climacteric product, its harvest in the suitable ripening stage is highly significant from a nutritional and economic perspective [11]

Methods

Results

Discussion

Conclusion