The Processing Algorithm Of Digital Television Signal Passive Bistatic Radar And Its Implementation On GPU

With the increasing number of low-speed and small UAVs,the problems brought about by it can not be ignored.Detection and early warning of low and slow targets before they bring danger to society is a serious task at present.Radar,as an effective method of target detection,has been developing for 80 years.With the increasingly fierce competition of electronic countermeasures,modern active radar faces many challenges.such as stealth targets,anti-radar missiles,low-altitude threats and electronic jamming,etc.They impede radar detection of targets.Therefore,the concept of external emitter radar is proposed.As a passive radar,it does not emit electromagnetic signals,but only receives the signals of existing electromagnetic waves reflected by targets in the air.The passivity makes it difficult for the enemy to find its position and to attack radar with missiles and electronic jamming.In this paper,digital TV signal is used as radar signal source of external radiator to detect small targets with low speed.Digital TV signal has wide bandwidth and large range resolution,which has great advantages for small target detection.In this paper,GPU is used as signal processing platform.As a high performance parallel computing device,GPU is widely used in deep learning,game and signal processing.External Radiator Radar No Signal Transmitter.The receiving part has two channels,receiving direct wave signal and echo signal respectively.The clutter jamming reflected by the jammer in the echo signal is much stronger than that of the target signal.Therefore,it is necessary to suppress the clutter interference in the echo signal by an adaptive filter.This paper compares the computational complexity and target detection performance of NLMS algorithm,RLS algorithm and ECA algorithm.The experimental results show that,on the one hand,the computational complexity of the NLMS algorithm is lower than that of the RLS algorithm and the ECA algorithm.At the same time,the ECA algorithm is more effective for the parallel implementation of the GPU as an open-loop algorithm.On the other hand,NLMS algorithm,RLS algorithm and ECA algorithm can achieve better clutter suppression.At the same time,for slow target,ECA algorithm is better than other two algorithms in target detection performance.Therefore,this paper comprehensively considers the target detection performance and computational complexity of the algorithm,and selects the ECA algorithm to implement clutter suppression and slow target detection.Since the ECA-B algorithm can obtain the same detection performance as the ECA algorithm,and the ECA-B algorithm can be processed in parallel on multi-GPU devices to improve the efficiency of signal processing,this paper mainly studies the external radiation source radar signal based on ECA-B algorithm.Handle parallel implementation methods on multi-GPU devices.Specifically,this paper studies the implementation of matrix basic operations on the GPU,and analyzes the methods of matrix multiplication and matrix inversion on the GPU.For the problem of matrix multiplication calling GPU global memory,the data reading speed is slow.Consider using shared memory to optimize to reduce the matrix multiplication running time.For the problem of matrix inversion computational complexity,Gaussian elimination method is used in this paper.Implementation,column-by-column normalization and elimination two-step operation complete the inversion;for the problem that the FFT operation in the distance-Doppler processing occupies a large computation space,consider reducing the FFT operation amount by the decimation operation,and doing the extraction before the extraction Low-pass filtering avoids spectral aliasing due to decimation.In addition,the GPU implementation and optimization of clutter suppression algorithm and distance-Doppler processing are carried out.Experiments show that the algorithm has a significant improvement in signal processing time on the GPU platform compared to the CPU.