Study Of Nonhomogeneous STAP And Its Application To Airborne Radar

Constructing new space-time adaptive processing (STAP) method that can overcome the weakness of the conventional statistical STAP is one of the main research directions of the current STAP techniques, whose purpose is to adapt to the practically nonhomogeneous environment. The premise of the superiority of the statistical STAP is to have ample training samples with interference that are independently and identically distributed (IID) with interference in the test sample, in order to estimate the covariance matrix. Samples that content with this condition are called homogeneous sample. However, in practical airborne radar system with space-time configuration, the degree of freedom of the system is very high, and the actual environment is often changeful, which cause the condition difficult to satisfy. At this time we call the training samples are nonhomogeneous. The performance of the conventional statistical STAP meets great degradation under nonhomogeneous samples or environment. This paper puts great emphasis on the investigation of STAP filtering and target detection in nonhomogeneous environment. After introducing some basic knowledge of nonhomogeneous STAP in Chapter 2, we primarily studied two complete nonhomogeneous STAP schemes: the integration STAP and the direct data domain algorithm.In Chapter 3, one of the method for sample choosing, nonhomogeneity detector (NHD) is investigated. To provide the foundation for choosing and analyzing the NHDs, a kind of performance index is needed. We propose a kind of performance index of NHDs, and make use of it to analyze the influence of interference target to the performance of NHDs. We find that a clutter-whitening NHD shows notable degradation when it uses a covariance matrix polluted by the interference target; and the identifiabilities of other NHDs are limited. An ideal NHD first should discover those stronger target signals whose influence to the covariance matrix can not be neglected; next it is best to be independent on the sample-estimated covariance matrix, in order to identify those weaker interference targets; finally the structure should be simple and convenient to realize. Simulations indicate that virtual NHDs suggested in this Chapter owe outstanding operation, whether interference target exists or not. Since on-lineinversion is not required, the performance is improved with acceptable computation load. A new solution to interference target, anti-interference-target STAP algorithm is proposed in chapter 4, which falls under the synthesis technique of sum pattern. Application of normal synthesis technique to the anti-interference-target STAP algorithm causes great loss of the suppression of the clutter. However, signal leaking anti-interference-target method results in smaller loss. Chapter 4 discusses various signal-leaking methods detailedly. Signal leaking produces loss in aperture, but performance improvement from its increase of the training samples actually remedy the loss of space-time aperture.In Chapter 5, the performance loss in the neighborhood of mainlobe clutter for the power selected training method (PST) in nonhomogeneous environment and its solution is studied first. PST algorithm nicely resolves the hollow nulling problem, but it makes overnulling more serious, which can be remedied by adaptive diagonal loading. The combination of PST and adaptive diagonal loading is of benefit to the suppression of clutter. Then, the influence is studied that every kind of nonhomogeneous phenomenon exists at the same time. Every kind of nonhomogeneous phenomenon, include the power fluctuation, the interference target and the discrete interference, exists at the same time in actual radar data, but many nonhomogeneous STAP methods can only solve a kind of these phenomenon. To reasonably use these methods together to get a practical nonhomogeneous STAP, analyzing the influence of coexist of the three kinds of nonhomogeneous phenomenon is demanded. Upon this foundation, the chapter finally examines with simulated data a practical integration S...