A Target Detection and Tracking System with Model-Based Clustering and Discrimination

The mission of the GMD system is to defend all 50 United States against a limited strike of Intercontinental-class Ballistic Missiles by adversaries from rogue nations. The system must perform detection, discrimination, battle management, and interceptfunctions, which require the integration of multiple sensors, communications, command and control, and weapon systems. Multi-sensor data fusion is a technique by which data from several sensors are combined to provide comprehensive and accurateinformation. In Phase I, we propose an advanced signal processing system suitable for sensor platforms such as XBR, SBIRS, UEWR and EKV seeker. Since a single feature detection is inadequate in dealing with the complex nature of radar images, especiallythose in high clutter environment, we propose to develop two independent feature detectors, namely, (1) statistical image model for feature detection and clustering in a high clutter environment, and (2) neural network-based unsupervised clustering forextraction of topological features. In particular, we will develop a prototype system with innovative technologies including (1) mixture statistical image model for feature detection in high clutter environment, (2) a new unsupervised neural network usedfor extraction of the topological spectral features, (3) multi-classifier fusion based on statistical evidence theory. The advanced signal processing system developed in Phase I and II has a great potential of commercial success. Although the technologydeveloped in this project is specific to the missile defense system, the overall architecture is applicable for many other government agencies (Air Force, Navy, DOT, NASA) and various industries. For example, the core technologies developed can be used inprocess industry where field data of heterogeneous sensors are collected in the central control room, displayed and analyzed.