Densely-Packed Target Data Fusion for Naval Mission-level Simulation Systems Navy SBIR FY2010.1 Sol No.: Navy SBIR FY2010.1 Topic No.: N101-101 Topic Title: Densely-Packed Target Data Fusion for Naval Mission-level Simulation Systems Proposal No.: N101-101-0850 Firm: 3 Phoenix, Inc. 14585 Avion Pwy Suite 200 Chantilly, Virginia 20151 Contact: Tushar Tank Phone: (919) 956-5333 Web Site: www.3phoenix.com Abstract: We propose a principled data fusion framework that is appropriate for an adaptive classifier implemented with supervised and multi-task learning. The detection and data fusion (DDF) engine will incorporate a novel Bayes-optimal multiple target tracking system. We will investigate several different metrics of the utility of data fusion in addressing strategic and tactical course of actions. We will perform testing on measured data to help define which is the most appropriate for Navy multi-sensor missions. In addition, we will develop new techniques for feature adaptation and selection based upon current operational scenarios within the battle space. Benefits: The proposed research has the opportunity to significantly advance the manner in which the Navy performs multi-sensor multi-platform data integration. The proposed algorithms will yield a high level of adaptivity, allowing for changing environmental, target and clutter conditions. This will be performed using a new class of supervised learning algorithms that allow the detection and data fusion (DDF) engine to adapt to sensing conditions. The proposed algorithms are appropriate for integration within supervised and multi-task learning, these exploiting a significant level of context. We also address the critical issue of multiple target tracking within this construct. In the proposed research we will build upon supervised and multi-task algorithms that are being developed by 3Pi under separate support. The proposed Phase I research will make the following specific contributions: (1) The proposed algorithms will take into account real-world sensing issues, such as the potential for noisy data/labels, as well as the cost of label/data acquisition. (2) A new Adaptive Kernel Elastic Net algorithm is proposed, which will provide a level of accuracy in feature selection that has previously been unavailable. (3) A novel Bayes-Optimal multiple target tracking system is proposed that will provide high level of accuracy in detection in dense target scenarios. The potential commercial applications of the research can be applied to security industry and imagining industry. The availability of multiple sensor data such as high spatial and spectral resolution satellite imagery has created a growth industry in land-use assessment, optimized natural resource extraction, habitat analysis, precision agriculture, and urban planning/infrastructure analysis. However, the variability of sensing parameters (i.e., cloud-cover, seasonal variations, incident and scattering angles relative to time of day and sensor orientation) can confound classification performance. Also the amount of training data is small. The algorithms developed under this effort could accommodate such high data volumes and adapt to sensor variability for improved image scene characterization. Such methods could be useful to the Bureau of Land management (BLM), the United States Geologic Service (USGS), the United States Department of Agriculture (USDA), and the intelligence community for the analysis of commercial, and non-commercial, remote sensing data. Another key application of the proposed research is in the area of security and surveillance. With increase in multi-spectral and visible imaging being deployed both in the commercial and defense space, a great volume of multi-platform data is being generated. Again we are faced with variability in sensing parameters and a dearth of training data for anomaly detection to guard against an asymmetric threat. The algorithms proposed would provide feedback to operators as to which features and are most informative and need attention. Return