Research On Radar Emitter Reconnaissance Processing Algorithm Based On GPU

Radar emitter reconnaissance is an important part of electronic countermeasures and a necessary condition for radar jamming and radar attack.As the electromagnetic environment of modern warfare becomes increasingly dense and complex,the amount and difficulty of information processing to be completed by radar emitter reconnaissance system are increasing day by day.How to accurately and quickly estimate the parameters of received radar pulses and realize signal sorting has become an intractable problem.The CUDA parallel computing based on GPU can greatly improve the computing efficiency and provide a new direction for the optimization of emitter reconnaissance processing algorithm.In this thesis,the emitter reconnaissance processing algorithm is studied,and its efficiency is optimized by GPU parallel computation.Specific work contents are as follows:1.A simulation system for reconnaissance and processing of radar emitter is built by adopting the heterogeneous platform of CPU+GPU and combining with software-based radar technology.2.Based on the CPU+GPU heterogeneous computing platform,the homogeneous digital channelization of the parallel optimized polyphase filter structure is realized.For non-cooperative signals,based on the uniform channelization of polyphase filter structure,the cross-channel signals are discriminated and the channels are reconstructed,and the non-uniform digital channelization of parallel optimization is realized.3.Based on the heterogeneous computing platform of CPU+GPU,the signal detection and parameter estimation of the received pulse signals are completed,and the pulse description words are formed.Signal detection algorithms include autocorrelation accumulation,multiple detection,non-coherent accumulation and constant false alarm detection.Parameters estimation includes time of arrival,carrier frequency,bandwidth,pulse width,pulse amplitude and angle of arrival estimation.DFT method and DFT-based Rife(Real-time Intermediate Flow Estimation)method are used for carrier frequency estimation,and correlation interferometer method is used for angle of arrival estimation.Both signal detection and parameter estimation are optimized in parallel,and their accuracy and acceleration effect are evaluated.4.Complete the design and implementation of signal sorting based on CPU+GPU heterogeneous computing platform.Signal sorting includes pre-sorting and main sorting.The density cluster sorting algorithm based on DBSCAN(Density-Based Spatial Clustering of Applications with Noise)was used for pre-sorting to realize the preliminary classification of pulse signals.In the pre-selection,each step of cluster sorting is analyzed,and the optimized part is accelerated in parallel.PRI(pulse recurrence interval)transform method was used for main sorting,and secondary separation of pulse signal was carried out.In the main sorting,the PRI transform method is modified to estimate the PRI value of the emitter pulse train with jitter PRI.