Implementation Of Airborne Early Warning Radar Signal Processing Method Based On GPU

Space-time adaptive processing of airborne early warning radar requires a certain number of independent and identically distributed samples to estimate the clutter covariance matrix,which is difficult to obtain in the scene of non-uniform clutter such as city and mountain area.The lack of sample numbers will lead to a sharp decline in the performance of clutter suppression,which will affect the target detection performance and tracking quality of the radar system.Therefore,there is an urgent need for algorithms which can suppress the clutter of airborne early warning radar effectively under the condition of heterogeneous clutter.In this context,knowledge-assisted STAP has received extensive attention.Knowledge-aided STAP acquires the prior information of scene by perceiving clutter which and reduces the influence of sample size on clutter covariance matrix estimation.Therefore,the knowledge-aided STAP can guarantee the performance of clutter suppression in a non-uniform environment.Airborne early warning radar has wide detection scope,complex scene and very large number of targets,so its signal and data processing has large amount of computation and high real-time requirement.The above factors synthetically determine that the signal processing system of airborne early warning radar requires a extremely strong computer power.GPU can meet the speed requirements of airborne radar signal processing because of its strong floating point computing power and powerful parallel ability.In recent years,the realization of high speed radar signal processing based on GPU platform has become an important research direction.This paper studys the problem of knowledge assisted space-time adaptive processing and acceleration method for airborne radar signal processing based on GPU platform.The main contents are as follows:1.In order to solve the problem of space-time adaptive processing in non-uniform environment,a knowledge assisted space-time adaptive processing method based on dynamic perception is studied.This algorithm can detect environmental information in real time to alleviate the mismatch between environmental knowledge and the actual environment.This algorithm firstly uses the orthogonal waveform to perceive the clutter environment.Then reconstructs the clutter covariance matrix by the obtained prior data.At last,a space-time adaptive filter is generated by using the reconstructed clutter covariance matrix to suppress clutter.The simulation experiment shows that the clutter suppression performance of the above KA-STAP algorithm is better than the 3DT method in the non-uniform environment.In this paper,the alternating direction multiplier algorithm(ADMM)is used to obtain prior information,which makes the time of solving the sparse reconstruction problem greatly shortened.2.Introduce the features of GPU and its advantages relative to CPU,and describes the CUDA architecture for GPU programming in detail from three aspects: programming model,execution model and hardware model.Finally,the mixed programming between GPU and MATLAB is introduced.3.The parallel implementation method of signal processing algorithm for airborne early warning radar based on GPU is briefly introduced,including the Simulation of airborne radar echo signal,matched filtering,clutter rejection and constant false alarm rate detection.In order to adapt to more clutter scenarios,the clutter suppression module not only implements the classic 3DT algorithm,but also realizes the knowledge aided space-time adaptive algorithm based on dynamic perception introduced in this paper.The experiment results show that GPU can markedly enhance the speed of radar signal processing over CPU serial computing.The data is larger,the improvement of speed is more remarkable.