Automatic Smoke and Obscurant Cloud Pattern Recognition from Visible and Thermal Imagery

U.S. Army smoke and obscurant field testing programs produce information and data for assessment of inventory and developmental obscurants. Determinations of obscurant cloud dimensions, locations, and centroids are often used for quantifying performance. To determine geometric cloud characterization parameters, current data reduction procedures require time-sensitive and potentially inconsistent manual (human) location of cloud perimeters from visible and infrared video imagery. The Phase I work plan proposes to identify both software and hardware components necessary to create an automated cloud perimeter location system. The plan seeks to integrate human cues into a knowledge base which will utilize image processing techniques and algorithms. The Phase I effort will also explore algorithms for the extraction of additional obscurant information, e.g., path radiance and cloud density, from the video imagery. Successful techniques and algorithms will be recommended for inclusion in the Phase II engineering of the perimeter detection system.