Chemical Imaging of Heterogeneous Muscle Foods Using Near-Infrared Hyperspectral Imaging in Transmission Mode.

Abstract

Foods and biomaterials are, in general, heterogeneous and it is often a challenge to obtain spectral data which are representative for the chemical composition and distribution. This paper presents a setup for near-infrared (NIR) transmission imaging where the samples are completely trans-illuminated, probing the entire sample. The system measures falling samples at high speed and consists of an NIR imaging scanner covering the spectral range 760-1040 nm and a powerful line light source. The investigated samples were rather big: whole pork bellies of thickness up to 5 cm, salmon fillets with skin, and 3 cm thick model samples of ground pork meat. Partial least square regression models for fat were developed for ground pork and salmon fillet with high correlations (R = 0.98 and R = 0.95, respectively). The regression models were applied at pixel level in the hyperspectral transmission images and resulted in images of fat distribution where also deeply embedded fat clearly contributed to the result. The results suggest that it is possible to use transmission imaging for rapid, nondestructive, and representative sampling of very heterogeneous foods. The proposed system is suitable for industrial use.