Performance Evaluation Modeling for Multi-Sensor ATR and Tracking System

We propose a new performance modeling and prediction framework for the evaluation of ATR and tracking systems with various sensor fusion algorithms. The primary goals are to enhance the operational ability of the ATR, to reduce the ambiguity of tracking closely spaced targets, and to develop a realistic performance prediction model. Using the Data Fusion Information Group information fusion model as a guide, we will develop a framework for system performance modeling and prediction which consists of five innovative components:1) New performance metrics for joint estimation and decision fusion solutions. 2) An image-level measure set which is very effective to predict system-level ATR performance. 3) A novel approach for online adaptation of the ATR and tracking parameters using the image characteristic measures and performance prediction. 4) A unified target identification and data fusion framework. It not only extends the multiple hypotheses tracking with efficient target identification algorithm, but also incorporates target features, attributes as well as target classification outputs from local data processors. 5) the usage of sensor management assures the improved accuracy of lower level fused data via the feedback from the higher level data fuser.