Automatic Battle Damage Assessment in Remotely Sensed Imagery

In the modern warfare, especially in the post cold war era, great effort has been devoted to precision strike to maximize damage on strategic targets while minimizing civilian casualties. Since the strike is limited to a small scale, accurate and rapid assessment of a strike's success is crucial in planning future attacks. The problem is complicated by numerous factors which influence the process such as image shift and variations due to small mobile objects and scattered debris. In this proposal, we describe a novel algorithm based on feature-level image registration and comparison for automatic damage assessment. This algorithm incorporates several state-of-the-art image processing techniques including Bayesian Pattern Theory, Flexible Template Matching (FTM) and Fractal Dimension Analysis. Specific Phase I effort will include (1) literature search, (2) acquisition of data for Automatic Battle Damage Assessment, (3) development of image segmentation and feature extraction software, (4) development of a damage assessment algorithm using Bayesian Pattern theory approach and FTM, (5) testing and evaluation of the algorithm, and (6) reports and Phase II recommendations. Phase II will include development of software demonstration systems and testing on sample imagery. Prof. Grenander and Dr. Srivastava, both of Brown University, will provide support in the area of feature extraction and pattern theory.