Detection And Parameter Estimation Of Maneuvering Target In Forward-looking Airborne Radar

The airborne phased array radar, which is mounted on a moving aircraft, is characterized by many advantages, such as wide range surveillance, long early warning time for low-altitude targets, etc. However, it is a great challenge to detect the modern maneuvering targets timely and effectively, which poses a serious threat to the future radar defense systems.The forward-looking configuration is common in airborne radar, however, the clutter is range dependent and the Independent and Identically Distributed(IID) training samples are finite,which would degrade the performance of detection. In this dissertation, the detection and parameter estimation of maneuvering target is studied, and the main conclusions have achieved are summarized as follows:Firstly, the estimation of clutter's Degrees of Freedom(DoF) under finite training samples was studied and an approach based on Random Matrix Theory(RMT) and Minimum Description Length(MDL) criteria was proposed. The effectiveness of the proposed method was analyzed theoretically, and the method provided theoretical basis for reduced rank Space-Time Adaptive Processing(STAP) in forward-looking airborne radar.Secondly, the impact of maneuvering degree to detection was analyzed, and the drawn conclusion was that the low degree of maneuvering can be ignored in the High Resolution Radar working at azimuth scanning mode. This conclusion provides a theoretical foundation for reducing the dimension of matched estimation. For the detection of low maneuvering target under partially homogeneous environments, an adaptive subspace whitening detection was developed, the analytical expressions of false alarm and detection probabilities were presented, and the Constant False Alarm Rate(CFAR) basic attribute was analyzed.Finally, a modified adaptive matched filter detector for high maneuvering target was proposed. In this detector, the Doppler frequency migration was compensated by the acceleration which was estimated separately, and the required training samples were reduced by diagonally loading technology. The effectiveness of the compensation and the asymptotically CFAR property was analyzed.