HYPERSPECTRAL IMAGING IN FRUIT QUALITY ASSESSMENT: SPECTRAL ANALYSIS, MACHINE LEARNING AND APPLICATIONS

Abstract

Fruit quality assessment is important for maintaining postharvest quality, reducing losses and improving the efficiency of sorting and grading operations. Conventional destructive methods provide reliable measurements of physicochemical, chemical, mechanical and microbiological attributes, but generally require sample preparation, laboratory analysis and destruction of the fruit. Non-destructive techniques such as visual inspection, machine vision, near-infrared spectroscopy, multispectral imaging, thermal imaging and electronic nose systems have been increasingly used for fruit quality evaluation, although their ability to provide detailed spatial and spectral information simultaneously is limited. Hyperspectral imaging (HSI) combines imaging and spectroscopy to acquire spatially resolved spectral information from intact fruits, allowing both internal quality attributes and external defects to be assessed. This review discusses the development and working principles of HSI, scanning and illumination modes, and the major steps involved in spectral data processing, including preprocessing, feature extraction, wavelength selection and dimensionality reduction. Applications of HSI for internal quality attributes such as soluble solids content, pH, firmness and total soluble solids, as well as external defects and fungal diseases, are reviewed with emphasis on spectral range, data analysis methods and machine-learning models. The major challenges include limited and cultivar-specific datasets, environmental variability, laboratory-to-field transferability, insufficient external validation, high-dimensional data and the lack of standardized protocols. Future research should focus on compact and portable HSI systems, real-time inspection, artificial intelligence, multimodal sensing, transfer learning, standardized datasets and commercial-scale validation to support reliable and practical fruit quality assessment.