Calibration and Correction for Non-Uniform Pixel Sensitivity in Digital NIR Imaging

Abstract

Digital near infrared (NIR) imaging of biological subjects is an increasingly important aspect of food quality inspection and bio-security in general. Digital NIR sensors, usually focal plane arrays, are known to have non-uniform sensitivity from pixel to pixel arising from detector manufacturing issues, operational conditions and design limitations. This study reviews the fundamentals for non-uniform pixel sensitivity and subsequently describes and tests a methodology for correcting such non-uniformities. The methodology assumes linear pixel sensitivity and employs a statistical analysis to distinguish between signal variation and pixel sensitivity. The system used to test the methodology employs a liquid crystal tunable filter (LCTF) based hyperspectral NIR imaging system capable of automatically acquiring a hyperspectral data cube in the 900 - 1700 nm range. Preliminary testing indicates promising results.