APPLE QUALITY ASSESSMENT IN THE ERA OF AI: NON-DESTRUCTIVE AND PREDICTIVE MODELING TECHNIQUES

Accurate apple quality prediction is crucial for ensuring postharvest efficiency, consumer satisfaction, and sustainable agricultural practices. Traditional methods based on manual inspection and destructive testing are often subjective and inefficient, prompting the need for automated, data-driven approaches. This review explores modern technologies applied to apple quality prediction, including hyperspectral and near-infrared spectroscopy, machine learning algorithms, deep learning models like convolutional neural networks (CNNs), and genomic prediction methods such as QTL mapping and genome wide selection (GWS). Recent studies have demonstrated the effectiveness of combining spectral data with intelligent optimization techniques like particle swarm optimization (PSO) for real-time assessment of internal traits such as sweetness, firmness, acidity, and defect detection. Additionally, integrating genomic data has enabled the prediction of inherited quality traits, supporting advanced breeding programs. The review synthesizes findings from over twenty recent studies to highlight progress, challenges, and future directions in this field. It recommends the development of portable, field- deployable tools, standardized evaluation protocols, and multidisciplinary collaboration. Ultimately, the convergence of AI, spectroscopy, and genomics is transforming apple quality monitoring into a scalable, intelligent system suitable for modern precision agriculture.