Inversion Of Optical Properties For Fruit Tissue Based On Hyperspectral Scattering Images

Optical properties of fruit are important factors affecting itself chemical composition, physical structure, physiological, pathological state. Measuring optical properties of the fruit has important significance for studying imaging of the internal structure of the organization, analyzing of chemical properties, physical structure and transmission characteristics within the organization, establishing detection and evaluation model of organization internal quality/state. In the study, diffuse approximation model combined with Monte Carlo simulation was used to study non-destructive measurement with the optical properties of the fruit tissue. Then, the optical properties of tomatoes tissue were extracted and classification/prediction model between optical properties and tomato ripeness was established. The main contents in this paper are as follows:In the study, diffusion model has a disadvantage that exist a larger deviations for reflectance when the detector is close to the light source. Therefore, the article proposed a moment-transformation method, which changes the form of raw data, then utilized subspace trust-region of interior-reflectance Newton method to inverse optical properties of fruit tissue. Compared to natural logarithm transformation, moment-transformation can decrease inversion deviations and increase the inversion accuracy of optical properties of fruit tissue. In certain case that SNR(Signal to Noise Ratio) is greater than 50 db, moment-transformation still obtain a higher inversion accuracy of optical properties.The hyperspectral imaging technology based on the steady spatial-resolved technology was used to measure the optical properties of tomatoes tissue, the classification model for tomatoes of 281 different ripeness were established combined with partial least squares discriminant analysis. Firstly, the diffusion model was used to extract the absorption coefficients and reduced scattering coefficients of ‘Sun Bright’ tomatoes different ripeness by hyperspectral imaging based on the steady spatially-resolved technology; then 10 random method was used to select model sample and partial least squares discriminant analysis algorithm was utilized to establish three classification(a total of 45 predicted samples) and six classification(a total of 67 predicted samples) model between optical properties and tomato ripeness. Partial least squares discriminant analysis(PLSDA) models yielded 92.1%, 84.4%, 92.3%, and 92.1% overall classification accuracies for three ripeness grades, when using the full spectra(500-950 nm) of,, and the effective attenuation coefficient(), respectively. am ’sm &’a sm m( )1 23 ’eff a a sm = ém m +m ù? ?The article studied prediction model of immature/mature tomatoes using optical properties. Hyperspectral scattering images were collected within 500-950 nm wavelength range using hyperspectral imaging device, the optical properties of tomatoes tissue was extracted by Farrell diffusion model, effective wavelength using mean sensitivity analysis were selected and prediction model of ‘Sun Bright’ tomatoes with support vector data description(SVDD) algorithm based on grid optimization technology was establish. KS algorithm selected 234 mature tomatoes for modeling from 240 samples and the remaining 40were used to prediction. SVDD models yielded 92.5%, 90.0% and 97.5% overall prediction accuracies for immature/mature grades, when using the full spectra(500-950nm) of,,), respectively. Compared to the full wavelength, the six effective wavelengths obtained 92.5%, 92.5% and 97.5% overall prediction accuracy. am ’sm’&a sm m...