Intelligent detection and waste control of hawthorn fruit based on ripening level using machine vision system and deep learning techniques

Abstract

Increasing marketability and waste management of agricultural products require quality assessment. Meanwhile, their marketability is largely affected by their shapes and overall appearance. Deep Learning (DL) has gained traction as a leading tool for computer vision tasks involving image detection and classification. This research was conducted in order to achieve better grading, reduce waste and at the same time increase the export and marketing of hawthorn fruit. In this regard, to classify images, 3 categorizes of hawthorn (unripe, ripe, and overripe) were acquired and the images were prepared using a well-designed illumination chamber. A data augmentation method was employed to improve the DL performance. After the pre-processing step, the capabilities of the suggested Inception-V3, ResNet-50, and the proposed DL models based on convolutional neural networks (CNN) used to grade the hawthorn fruit. In comparison with other methods, the Inception-V3 surpassed the overall validation accuracy of 100%, indicating superiority of this network over the other classifiers. Therefore, CNN and image processing techniques can be effective in increasing marketability, controlling waste and improving traditional methods used for grading hawthorn fruit.

Keywords

• Fruit classification
• Waste management
• Machine learning
• Convolutional neural networks
• Data augmentation