Orange Quality Grading with Deep Learning

2025-03-27

AI Summary

To address incomplete peel-defect detection and poor grading robustness caused by single-view imaging in automated citrus sorting, this paper proposes a multi-view image stitching and deep learning–integrated method for external quality grading of oranges. We introduce a novel single-fruit multi-view image stitching model to overcome the limitations of conventional single-perspective approaches. The method employs a dual-CNN architecture—ResNet-18 and SqueezeNet—combined with multi-view synthesis, comprehensive data augmentation, and end-to-end supervised training. Experimental evaluation on real-world production-line images achieves 96.2% classification accuracy, significantly outperforming single-view baselines. The system enables fully automated three-tier classification (“good,” “defective,” and “undefined”), balancing grading accuracy, processing efficiency, and industrial deployability. This work advances practical, vision-based quality control for high-throughput citrus postharvest automation.

Technology Category

• Machine Learning: Multi-instance/Multi-view Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Orange grading is a crucial step in the fruit industry, as it helps to sort oranges according to different criteria such as size, quality, ripeness, and health condition, ensuring safety for human consumption and better price allocation and client satisfaction. Automated grading enables faster processing, precision, and reduced human labor. In this paper, we implement a deep learning-based solution for orange grading via machine vision. Unlike typical grading systems that analyze fruits from a single view, we capture multiview images of each single orange in order to enable a richer representation. Afterwards, we compose the acquired images into one collage. This enables the analysis of the whole orange skin. We train a convolutional neural network (CNN) on the composed images to grade the oranges into three classes, namely good, bad, and undefined. We also evaluate the performance with two different CNNs (ResNet-18 and SqueezeNet). We show experimentally that multi-view grading is superior to single view grading.

Problem

• Automated orange grading using deep learning for quality classification
• Multiview image analysis improves accuracy over single-view systems
• Comparison of CNN models (ResNet-18, SqueezeNet) for grading performance

Innovation

• Multiview image capture for richer representation
• Image collage for whole orange skin analysis
• CNN training on composed images for grading