Hyperspectral target tracking

We introduce a non-kinematic based approach to autonomous target tracking using a new set of Independent and Indirectly Generated Attributes (IIGA) from hyperspectral imagery. The IIGA method addresses the detection of rare signal appearance (i.e., targets represented by a few pixels), which is often the case in remote sensing target tracking. The proposed method demonstrates that object distinctness can be preserved, or perhaps accentuated, by contrasting hyperspectral samples, indirectly, through differences between each sample and a series of unrelated random samples in order to generate new attribute sets. Object distinctness is captured by features of the new attribute sets´ underlying distributions. Preliminary results are promising using a small but challenging dataset.