 |
Center for Nonlinear Studies |
Current detection methods for solid materials can suffer high false alarm rates. Reduction of the number of false alarms may be possible with a hyperspectral representation of solids that has less variability. To reduce this hyperspectral variability we characterize solid materials through spectral features derived using Gaussian basis functions and spectral fitting with regularized regression. Our goal is to improve compositional exploitation in an LWIR hyperspectral imaging sensor data cube using a detection algorithm based on spectral features. There are material-specific properties of these spectral features that show subtle variability when considering changes in morphology or measurement conditions. This new approach has good initial detection results across material particle size, measurement angle, and atmospheric conditions using LLNL experimental measurements for validation purposes. To help those unfamiliar with this research topic, the talk provides an introduction to infrared theory and hyperspectral data analysis.
Host: James Theiler