| Objective: ||Develop real-world performance estimation models, evaluation techniques, and fusion models for multimodal sensing to anticipate, find, fix, track, understand, and assess objects of interest.
|| Description: ||The Air Force Research Laboratory (AFRL) is assessing different sensor fusion/ATE and tracking techniques that could be fielded for operational use. To address the questions of whether or not the systems are useful to transition, it is paramount to determine the expected performance of the ATE and/or tracking systems over operational conditions. The expected performance can be gleamed through theoretical, simulated, or empirical analysis; but the best approach is the union of all three methods in a unified performance model (PM). Exercising ATE and/or tracking systems on operational data can give a real-world regression assessment of what is capable. Simulated results can expand the operational performance prediction over emerging data. Finally, theoretical results would afford curves/surfaces of performance that determine whether figures of merit can be achieved. The goal is to start with operational scenarios and determine what sensors, systems, and processes are being used or planned to be used and determine the expected performance capabilities of the exploitation. The offerer should address either the sensor fusion/ATE area or the area of tracker performance to provide the best resource management utilization for tracking targets in a single pass. The first area of interest would require a 1) a sound operational scenario that includes deployed or emerging sensors and data extraction techniques (i.e., formats and ancillary information), 2) a detailed ATE methodology that supports the data exploitation, and 3) a design of experiments demonstration using collected and simulated data from which to derive a theoretical PM. The optimal proposed strategy would be one that includes experience with deployed systems, real-world data analysis, performance documentation, and theoretical justifications of performance. The latter area of interest would include 1) developing mechanisms for the detection and characterization of tracker data association ambiguity, and 2) developing methods for representing the ambiguity/performance prediction to a sensor resource manager such that sensor asset rerouting or rescheduling can be accomplished to potentially resolve the tracking ambiguities. The project is intended to support sensor management techniques which require a model of the combined sensor/automatic target recognition (ATR) techniques performance prediction models or the tracking models. Each respondent should concentrate on either the ATE research or the tracking problem. Phase I production include summaries of various methods, experiments, metrics, and performance results.
|| ||PHASE I: Develop an operational automated exploitation performance prediction and/or tracking performance prediction and estimation strategy with appropriate evaluation criteria. Design, implement, and test the system over data. Deliver methodology, data used, test results, interpretation of results and recommendations for further work to AFRL.
|| || ||PHASE II: Extend the system with a comprehensive set of metrics, information needs, and theoretical performance capabilities over various operating conditions and domain experts (e.g., image analysts). Products would include delivered software design and effective use of methods over collected data.
|| ||DUAL USE COMMERCIALIZATION: Military application: The techniques developed under this SBIR would be useful for supplying performance models to train a variety of algorithms for antiterrorism missions.
Commercial application: Domains that could benefit from automatic target exploitation performance modeling and tracking performance modeling include 1) medical diagnosis, 2) quality inspection, and 3) disaster assessment.
|| References: ||1. Sensor exploitation technology of particular interest to AFRL/SNA may be seen in the several dozen papers at: http://www.mbvlab.wpafb.af.mil/paper.html.
|Keywords: ||ATR, tracking|