Visualization of lamb adulteration based on hyperspectral imaging for non-destructive quantitative detection

Abstract

Objective: This study aimed to establisha rapid and accurate prediction method of lamb adulteration by using visible/near-infrared （400~1 000 nm） and short-wave near-infrared （900~1 700 nm） hyperspectral imaging techniques. Methods: The data acquisition of lamb adulterated with different proportions of duck meat using visible/near-infrared （400~1 000 nm） and short-wave near-infrared （900~1 700 nm） hyperspectral imagers was performed to compare the effect of partial least squares （PLS） modeling with different spectral preprocessing methods in the two band ranges. Then the normalized preprocessing method was selected in the visible-NIR band, and the standard normal variate transformation （SNV） preprocessing method was selected in the short-wave infrared band. After the optimal preprocessing of the spectral data on to the two bands separately, the feature wavelengths were selected using the successive projections algorithm （SPA）, the competitive adaptive reweighted sampling （CARS）, the Interval random frog （iRF） and the Synergy intervals PLS （SiPLS）. Results: The best lamb adulteration prediction using SNV-SPA-PLS model in the short-wave near-infrared （900~1 700 nm） bands, was achieved, and with the prediction set model evaluation coefficients of R2p=0.968 4, RMSEP=0.058 2, RPD=5.625 4, relaiable image inversion results were obtained. Conclusion: The rapid and nondestructive quantitative detection of lamb adulteration can be achieved by using hyperspectral imaging techniques in different wavebands.

Keywords

• lamb
• duck meat
• adulteration
• hyperspectral
• quantitative detection
• visible/near-infrared
• short-wave near-infrared