Citrus fruit classification
Pro Research Analysisby 
Searched over 200M research papers
Citrus Fruit Classification: Taxonomy and Horticultural Groups
Citrus fruits are globally significant, with major producers including China, Brazil, and India. The main horticultural groups are sweet oranges, mandarins, grapefruits, pummelos, and common acid members like lemons and limes. These groups are distinguished by morpho-taxonomic traits, such as size, shape, and color, and include hundreds of cultivars, many of which are the result of both natural selection and intentional breeding efforts Singh2021Peña2022. Molecular marker studies have clarified that most cultivated citrus types derive from just a few true species, despite the wide variety seen in markets Moore2001Peña2022.
Traditional and Manual Citrus Fruit Grading
Historically, citrus fruit grading has relied on human visual inspection, which can lead to inconsistencies and errors in sorting. Manual grading is based on external features like size, shape, and color, but maintaining uniformity is challenging Louis2020Varjão2019. This has driven the need for more reliable, automated classification systems.
Computer Vision and Machine Learning in Citrus Classification
Recent advances have focused on automating citrus fruit classification using computer vision and machine learning. Early systems used color-based selection and were implemented on embedded devices like Raspberry Pi, but these often required significant computational resources . More sophisticated approaches use image acquisition, preprocessing, and feature extraction, followed by machine learning models such as Support Vector Machines (SVM), achieving high classification accuracy (up to 97.3%) .
Deep Learning and Hybrid Models for Citrus Fruit Classification
Deep learning, especially convolutional neural networks (CNNs), has greatly improved the accuracy and speed of citrus fruit classification. CNN-based models can distinguish between citrus types with similar color and shape, achieving accuracy rates as high as 96% . Hybrid models, such as combining ResNet101 for feature extraction with SVM for classification, have further increased performance, reaching accuracy rates of 99.81% and excelling in sensitivity, specificity, and precision . These models are particularly valuable for food processing and quality control applications.
Disease and Defect Detection in Citrus Fruits
Beyond type classification, deep learning models are also used to detect diseases and physical defects in citrus fruits. Custom CNN architectures and transfer learning approaches have been developed to classify both healthy and diseased fruits, as well as to identify specific diseases like Alternaria alternata and Thrips. These models can achieve high accuracy (up to 95.7% for disease classification and 99% average precision for disease detection), helping to reduce post-harvest losses and improve fruit quality Rehman2021Uğuz2022.
Conclusion
Citrus fruit classification has evolved from manual inspection to advanced computer vision and deep learning techniques. Modern systems can accurately classify fruit types, detect diseases, and identify defects, supporting the citrus industry in quality control and reducing economic losses. The integration of machine learning and deep learning models continues to enhance the efficiency and reliability of citrus fruit classification across the supply chain Louis2020Varjão2019Behera2024+3 MORE.
Sources and full results
Most relevant research papers on this topic