ASABE Technical Library - Abstract

Abstract. This study used a line-scan high-throughput Raman imaging system to authenticate milk powder. A 5 W 785 nm line laser (240 mm long and 1 mm wide) was used as a Raman excitation source. The system was used to acquire hyperspectral Raman images in a wavenumber range of 103–2881 cm⁻¹ from the skim milk powder mixed with two nitrogen-rich adulterants (i.e., melamine and urea) at eight concentrations from 50 to 10,000 ppm. An acoustic mixer that utilizes high-intensity acoustic waves was used to prepare milk-adulterant mixtures. The mixed samples were put in sample holders with a surface area of 150 mmx100 mm and a depth of 2 mm for push-broom image acquisition. Varying fluorescence signals from the milk powder were removed using a correction method based on adaptive iteratively reweighted penalized least squares. Image classifications were conducted using a simple thresholding method applied to single-band fluorescence-corrected images at unique Raman peaks selected for melamine (673 cm⁻¹) and urea (1009 cm⁻¹). Chemical images were generated by combining individual binary images of melamine and urea to visualize identification, spatial distribution, and morphological features of the two adulterant particles in the milk powder. Limits of detection for both melamine and urea were estimated in the order of 50 ppm.