Hyperspectral and 3D point cloud imaging unraveling of physio-chemical shifts during drying of purple carrot and golden kiwifruit

Quality dried food production remains challenging due to limitations in processing and preservation. Quality driven process optimization is essentially relaying on real-time monitoring. Current study investigate non-invasive optical and invasive analytical techniques for quality analysis during drying of purple carrots and golden kiwifruit. Variations in moisture content (MC), carotenoids, anthocyanin, lutein, vitamin-C, shrinkage and color characteristics were analyzed using invasive methods and then compared with hyperspectral imaging (HSI) and 3D point cloud-based predictions. Spectral information (400–1002 nm) was correlated with several predictive models, with Gaussian Process Regression (GPR) and Gradient Boost Regression (GBR) yielding the most accurate results. The models successfully predicted MC (R p 2 = 0.965, RMSEP = 0.0503), anthocyanin (R p 2 = 0.875, RMSEP = 0.200), and β-carotene (R p 2 = 0.838, RMSEP = 0.0789) for purple carrots and MC (R p 2 = 0.989, RMSEP = 3.0436), β-carotene (R p 2 = 0.896, RMSEP = 0.1275), lutein (R p 2 = 0.755, RMSEP = 0.4093), and vitamin C (R p 2 = 0.947, RMSEP = 5.2349) for golden kiwifruit. Furthermore, for purple carrots, HSI-based total color change (ΔE) showed a higher correlation with colorimeter data (r 2 = 0.91) than 3D point cloud data (r 2 = 0.84). However, for golden kiwifruit, 3D point cloud-based ΔE has a stronger correlation (r 2 = 0.92) than HSI data (r 2 = 0.78) which are in-line with Bland-Altman analyses. The findings suggest that simpler, cost-effective and computationally efficient imaging sensors like multispectral imaging (MSI) can effectively replace HSI, paving the way for the development of smart dryers with real-time quality monitoring. • Accurate prediction of quality attributes alongside moisture content using Hyperspectral data. • Integrated 3D scanning and HSI data tracks spatiotemporal optical changes, closely correlating with color and shrinkage data. • Moisture independent pre-processed spectral data gives high prediction accuracy for quality components. • Assessment of quality shift during drying of purple carrots and golden kiwi at different temperatures. • Multivariate analysis supports real-time non-invasive monitoring of quality shift during drying.