Joint Target Tracking and Classification using GMTI/HRR Data

"Target tracking and classification form a coupled problem due to mutual dependence of the continuous kinematic state and discrete target class. Many tracking systems treat tracking and classification either separately or only utilize the one-way couplingbetween tracking and classification. We propose to develop new algorithms and software to perform joint tracking and classification (JTC) using GMTI sensor measurements and features derived from high range resolution (HRR) profiles by exploiting thecoupling between the two systems. We shall implement the JTC algorithms in ALPHATECH's PFCT tracker, a multi-target multi-sensor multiple hypothesis tracker to demonstrate the feasibility of the concept. The range profile for a target is a function of theorientation (aspect and depression angles), range, disturbances in the propagation medium, and signal processing. In order to mitigate the presence of noise in range profiles, we propose to perform robust classification using features extracted from theincoming and stored range profiles using information theoretic feature selection methods, and multi-layer perceptron neural network (MLPNN) based feature prediction models. We shall demonstrate improved measures of performance by JTC using simulated andreal GMTI data from AMSTE I and II and real HRR data from MTFP. Development of a joint tracking and classification system has a wide range of military and commercial applications. This resear