New research in China suggests that combining traditional color cameras with near-infrared imaging could make automated citrus grading systems much better at detecting surface defects. The technology has the potential to improve packhouse efficiency, reduce the number of defective fruit reaching consumers and help growers maximize the value of their crop.
The study, published in the journal Agriculture, was conducted by Jingxi Luo, Tao Wen and Dapeng Li, all from Central South University of Forestry and Technology; and Zhanwei Yang and Ying Cao from Central South University of Forestry and Technology and Kunming University of Science and Technology.
Surface appearance is one of the biggest factors determining citrus quality and market value. Fruit with blemishes, disease symptoms or cracks often receives lower grades and may reduce storage life. While many commercial sorting systems rely on standard color cameras, those cameras can struggle to detect certain defects or distinguish between similar blemishes.
To address that challenge, the researchers developed an artificial intelligence (AI) system that combines images from both visible light and near-infrared cameras. Visible-light images capture color differences, while near-infrared images highlight subtle changes in fruit texture and defect boundaries that are difficult to detect with the human eye. By combining information from both types of images, the new system was able to identify defects more accurately than systems using either imaging method alone.
The research team collected more than 1,600 images of citrus fruit with common defects, including citrus canker, pest damage, melanose and cracks. After image enhancement, the dataset expanded to more than 2,500 paired images that were used to train and test the AI model.
The new system consistently outperformed existing AI models. It achieved higher precision, recall and overall detection accuracy while requiring no additional model size, making it practical for future commercial grading equipment. The researchers found it was especially effective at detecting small defects, identifying blemishes of different shapes and reducing false detections caused by background color variations.
The study points toward improvements that could benefit the citrus industry after harvest. More accurate automated grading could help packinghouses sort fruit more consistently, reduce labor demands and ensure higher-quality fruit reaches premium markets. Better defect detection may also improve traceability and reduce losses caused by fruit with hidden or difficult-to-see surface damage.
The researchers note that additional work is needed before the technology is widely adopted commercially, including testing under real packinghouse conditions and expanding the system to recognize additional citrus varieties and defect types. Still, the results demonstrate how advances in AI and imaging technology could play an increasingly important role in citrus quality control.
Read the full study here.
Share this Post