Research On Space-time Adaptive Processing Algorithm For Sky-wave Radar

The sky-wave radar has the capability of long-range target detection and early warning,which plays an important role in national defense and civil fields,and is highly valued by world powers.Sea clutter suppression is a key problem in the sky-wave radar.Because space-time adaptive processing(STAP)is an effective clutter suppression technique,it is of great significance to study the STAP in sky-wave radar.However,STAP will face many challenges in the practical sky-wave radar clutter application,such as high demand for training samples,non-homogeneous clutter environment and heavy computation,and its sea clutter suppression performance will be degraded.Aiming at these problems,this dissertation studies sky-wave radar STAP algorithm from three aspects: training samples selection,clutter covariance matrix reconstruction and reduceddimension.The main work of this dissertation is summarized as follows:(1)Aimed at non-homogeneous clutter environment,this dissertation proposed a training samples selection method based on system identification for sky-wave radar STAP,it can obtain more training samples and get better performance of sea clutter suppression.We noticed that two signals with completely different waveforms may have the same covariance matrix.The traditional method based on waveform similarity only selects samples with similar waveform,which results in the loss of many valid samples.Based on this consideration,we propose to use the echo state neural network to judge whether the candidate samples have the same covariance matrix with the clutter of the cell under test(CUT),so that the samples with the same clutter covariance matrix are selected as training samples in the proposed method.(2)To solve the problem of insufficient training samples,this dissertation proposed a sky-wave radar STAP method based on covariance matrix reconstruction.This method makes full use of the clutter information of the CUT to improve the estimation accuracy of clutter covariance matrix under training samples shortage.We noticed that when estimating clutter covariance matrix,the traditional method based on the statistic information of training samples discards the CUT that might have the target echo.However,the CUT itself contains accurate clutter information.Based on this consideration,we propose to reconstruct the clutter covariance matrix based on the CUT data.In addition,according to the training samples,modify the reconstruction to eliminate the target signal and avoid the target cancellation.(3)To solve the problem of heavy computation in full dimension STAP,this dissertation proposed a sky-wave radar reduced-dimension STAP method based on sparse filtering.This method can adaptively select the auxiliary channels and improve the clutter suppression performance.Unlike the traditional fixed-structure reduced-dimension methods such as JDL,the proposed method is based on sparse theory and can adaptively select the appropriate auxiliary channels to construct the transform matrix for dimensionality reduction.Considering that sea clutter of the sky-wave radar is low-rank,we can effectively suppress sea clutter by selecting a small number of auxiliary channels with better performance to construct the transform matrix.When designing the STAP filter,more coefficients of the weight will be shrunk to zero by imposing a sparse regularization to adaptively select the auxiliary channels.