A morphological adaptation to spectral target detection searches

Abstract

Target detection results are commonly very difficult to visualize once completed – many small points are presented which may be cluttered or so dispersed that it may be challenging to find them visually. This work adds a spatial component to the spectral detection through the addition of a spectral-based morphological search. The results improve visualization by providing polygons at the target level rather than point designations at the pixel level. While this work was conducted on multispectral imagery, it is also valid for other imagery domains.