In this work, BTG developed the technology that allows one to identify and characterize fast events. In real time, we dynamically process hyperspectral information of a scene, specifically analyzing its temporal behavior. The goal is to detect fast and super-fast events like explosions, fast-moving objects and instant changes in chemical composition of the air and other materials.
Hyperspectral images of reactions
Until recently, the enormous amounts of hyperspectral information confined us to static hyperspectral data processing. Hyperspectral techniques were used for finding certain objects, chemicals, or anomalies in a picture, frame by frame, statically. Dynamic (temporal) analysis was developed primarily for astrophysical applications performed a long time after the frames had been captured.
In this work, we took advantage of emerging hardware technologies that allow one to look at hyperspectral information dynamically: by characterizing temporal changes as they occur. We apply methods from astrophysics (supernova observations) and present our unique algorithms for contemporaneous dynamical analysis of hyperspectral data.
The first application addresses the question: have there been any sudden changes in the hyperspectral pattern of a scene? If there were sudden changes, were those changes related to a substantial energy release? These questions do not depend on assumptions about specific spectral patterns, chemical composition, or shapes: we look for any changes in a scene. Such dynamical analysis can therefore offer a unique opportunity to react promptly to fast events without prior knowledge about what was to occur.