Context-aided tracking with an adaptive hyperspectral sensor

A methodology for the context-aided-tracking of ground vehicles in remote airborne imagery is presented in which a background model is inferred from hyperspectral imagery. The functional materials in a scene are determined and become a background model. Here, a manual method of forming the model is presented, as well as a novel autonomous method which exploits the emerging class of adaptive multiple-object-spectrometer instruments. A multiple-hypothesis-tracker is introduced, which relies on background statistics to form track costs and associated track maintenance thresholds. These statistics include detectability, track density, and false measurement density. Traditionally, these statistics are uniform constants, but the advent of the background model allows for spatially-varying statistics. The context-aided-tracker, which uses these statistics, is shown to achieve an increase in performance.