HYPERSPECTRAL IMAGING-BASED SPATIALLY-RESOLVED TECHNIQUE FOR ACCURATE MEASUREMENT OF THE OPTICAL PROPERTIES OF HORTICULTRAL PRODUCTS

HYPERSPECTRAL IMAGING-BASED SPATIALLY-RESOLVED TECHNIQUE FOR ACCURATE MEASUREMENT OF THE OPTICAL PROPERTIES OF HORTICULTRAL PRODUCTS By Haiyan Cen Hyperspectral imaging-based spatially-resolved technique is promising for determining the optical properties and quality attributes of horticultural and food products. However, considerable challenges still exist for accurate determination of spectral absorption and scattering properties from intact horticultural products. The objective of this research was, therefore, to develop and optimize hyperspectral imaging-based spatially-resolved technique for accurate measurement of the optical properties of horticultural products. Monte Carlo (MC) simulations and experiments for model samples of known optical properties were performed to optimize the inverse algorithm of a single-layer diffusion model and the optical designs, for extracting the absorption (μa) and reduced scattering (μs') coefficients from spatially-resolved reflectance profiles. The logarithm and integral data transformation and the relative weighting methods were found to greatly improve the parameter estimation accuracy with the relative errors of 10.4%, 10.7%, and 11.4% for μa, and 6.6%, 7.0%, and 7.1% for μs', respectively. More accurate measurements of optical properties were obtained when the light beam was of Gaussian type with the diameter of less than 1 mm, and the minimum and maximum source-detector distances were 1.5 mm and 10-20 transport mean free paths, respectively. An optical property measuring prototype was built based on the optimization results. The instrument was used to measure the optical properties, and assess quality/maturity, of 500 ‘Redstar’ peaches and 1039 ‘Golden Delicious’ (GD) and 1040 ‘Delicious’ (RD) apples. A separate study was also conducted on confocal laser scanning and scanning electron microscopic image analysis and compression test of fruit tissue specimens to measure the structural and mechanical properties of GD and ‘Granny Smith’ (GS) apples under accelerated softening at high temperature (22 oC)/high humidity (95%) for up to 30 days. The absorption spectra of peach and apple fruit were featured with the absorption peaks of major pigments (i.e., chlorophylls and anthocyanin) and water, while the reduced scattering coefficient generally decreased with the increase of wavelength. Partial least squares regression resulted in various levels of correlation of μa and μs' with the firmness, soluble solids content, and skin and flesh color parameters of peaches (r = 0.204-0.855) and apples (r = 0.460-0.885), and the combination of the two optical parameters generally gave higher correlations (up to 0.893). The mean value of μa and μs' for GD and GS apples for each storage date was positively correlated with acoustic/impact firmness, Young’s modulus, and cell parameters (r = 0.585-0.948 for GD and r = 0.292-0.993 for GS). A two-layer diffusion model for determining the optical properties of fruit skin and flesh was further investigated through solid model samples. The average errors of determining two and four optical parameters were 6.8% and 15.3%, respectively, for MC reflectance data. The errors of determining the first layer of model samples were approximately 23.0% for μa and 18.4% for μs', indicating the difficulty and also potential in applying the two-layer diffusion model for fruit. This research has demonstrated the usefulness of hyperspectral imaging-based spatiallyresolved technique for determining the optical properties and maturity/quality of fruits. However, further research is needed to reduce measurement variability or error caused by irregular or rough surface of fruit and the presence of fruit skin, and apply the technique to other foods and biological materials.