Hyperspectral near infrared image calibration and regression

Area of Science:

• Optical Imaging
• Spectroscopy
• Data Analysis

Background:

• Reference materials are crucial for converting camera signals to reflectance and absorbance in diffuse reflectance imaging.
• Traditional white and dark references can be enhanced with additional targets to improve transformation accuracy.

Purpose of the Study:

• To provide an overview of hyperspectral image regression.
• To assess the impact of reflectance calibration on image interpretation using partial least squares regression.

Main Methods:

• Investigated linear and quadratic reflectance transformations using additional reflectance targets.
• Applied partial least squares regression for hyperspectral image analysis.
• Evaluated measurement and prediction errors for different transformation methods.

Main Results:

• Linear and quadratic transformations with extra targets decreased average measurement errors.
• These transformations facilitated model pseudorank estimation in image regression.
• Column and wavelength-specific quadratic transformations yielded the lowest errors, preserving spatial information and improving concentration maps.

Conclusions:

• Enhanced reflectance calibration using additional targets improves accuracy in diffuse reflectance imaging.
• Advanced transformations are essential for precise hyperspectral image interpretation and quantitative analysis.
• Optimized calibration strategies are key to reducing errors in concentration mapping.