Algorithm Development for Predictive Battlespace Awareness

UltraHiNet, LLC (UHN) proposes to develop innovative high-performance computational strategies for real-time (RT) CSAR auto-mated target recognition (ATR) and coherent change detection (CCD). These strategies will enable sustained tera to petaflop computational rates for CSAR and will support very high I/O rates to meet the real-time (RT) demands for staring CSAR observation of areas up to 20KM in diameter using architectures that meet the size, weight, power and other constraints of on-board processing systems. We propose to lev-erage our existing expertise for developing and mapping image and signal processing algorithms to parallel RT architectures (CSPARTA), to achieve the following Phase 1 objectives: (a) Analysis of GFI (government furnished information) CSAR algo-rithms as well as of application specific data collection rates to determine key parameters including operation mix, data lay-out, I/O requirements, and computational accuracy; (b) Mapping of GFI CSAR algorithms to parallel RT architectures (e.g., multicore, GPU, FPGA, clusters of such processors, etc.) that meet the size, weight, and power constraints of an on-board processing system; (c) Performance analysis of parallel RT implementations of the CSAR algorithms, to include computational and I/O cost, as well as error propagation and accumulation; and provide prototype imple-mentation on one multicore processor. BENEFIT: The proposed research will constitute a breakthrough in the solu-tion of problems related to improved performance for efficient, accu-rate image reconstruction from multiple views. These problems occur extensively throughout both the military and commercial sectors: the potential payoff is high. UHN will license or sell the solution to large aerospace and avionics companies for military applications. In addition, we plan to collaborate with commercial companies involved in law enforcement technologies, environmental monitoring and medical imaging applications.