Detecting and identifying weak gaseous plumes using thermal imaging data is complicated by many factors. There are several methods currently being used to detect plumes. They can be grouped into two categories: those that use a chemical spectral library and those that don't. The approaches that use chemical libraries include physics-based least squares methods (matched filter). They are "optimal" only if the plume chemical is actually in the search library but risk missing chemicals not in the library. The methods that don't use a chemical spectral library are based on a statistical or data analytical transformation applied to the data. These include principle components, independent components, entropy, Fourier transform, and others. These methods do not explicitly take advantage of the physics of the signal formulation process and therefore don't exploit all available information in the data. This paper describes generalized least squares detection using gas spectra, presents a new detection method using basis vectors, and compares detection images resulting from applying both methods to synthetic hyperspectral data.
Keywords: LWIR; Plumes; basis vectors; detection; generalized least squares.