Research On Real-Time Signal And Data Processing Of Passive Radar Based On High-Performance Computing Architecture

With the development of electronic countermeasures technology,traditional radars are facing significant challenges in terms of anti-interference,anti-stealth,anti-radiation destruction,and anti-low altitude penetration capabilities.External radiation source radars have emerged as a hot research field both domestically and internationally in recent years due to their low frequency operation and passive detection characteristics,which provide unique advantages in the "four countermeasures".However,the target echo signal of external radiation source radar is extremely weak and usually concealed under the background of strong direct waves and multipath interference.This requires large-scale data processing with high computational complexity,which poses a serious challenge to the real-time implementation of external radiation signal and data processing.In response to this challenge,this article proposes a real-time detection and tracking technology for external radiation source radar targets using a high-performance architecture of GPU+CPU+PCIE.This approach combines the high concurrency capability of the GPU,the complex logic processing capability of the central processing unit,and the high-speed data transmission capability of the high-speed serial computer extension bus standard interface.The specific research content is as follows:First,based on an analysis of the detection principle and signal processing flow of external emitter radar,a real-time signal processing system and software architecture for external emitter radar are designed using high-performance architecture.To address the problem of high data throughput in multi-channel external radiation source radar systems,a high-bandwidth PCIE3.0×8 is used as the transmission interface,and a fast storage method of cyclic storage queue is adopted,effectively reducing the time consumed by data transmission.In terms of software architecture design,the system is decoupled and subdivided into small modules to achieve modular processing,from received signal preprocessing to target detection and tracking.This design approach facilitates system upgrades and maintenance.In terms of parallel processing for clutter interference suppression,this paper proposes an extended clutter cancellation batch inter-segment parallel algorithm based on LDLT decomposition to address the time-consuming problem of matrix inversion.Firstly,based on GPU multithreading parallel processing technology and combined with the same submodule characteristics of each segment of the ECA-B algorithm,the efficiency of the ECA-B algorithm is improved through segmented parallel processing.Then,to address the time-consuming problem of data transmission in the traditional ECA-B algorithm inversion process,a parallel iterative inversion method based on LDLT is proposed using the conjugate symmetry property of the autocorrelation matrix.The inversion process is achieved through two CUDA kernel functions,which save data transmission time in the matrix inversion process and further improve the efficiency of inter-segment parallel algorithms.The experimental results show that compared with traditional algorithms,the algorithm proposed in this paper has higher timeliness and effectiveness.In terms of parallel processing for object detection and tracking,a method for parallel processing is proposed to construct a delay sliding matrix and perform zero filling operations.This method addresses the problem of computational redundancy in the process of constructing the delay sliding matrix in range Doppler processing,as well as the need for separate zero filling operations in data processing.This method significantly reduces the algorithm’s time consumption.Additionally,a constant false alarm detection method based on protocol summation is proposed,which effectively shortens the solving time of detection coefficients and improves the efficiency of constant false alarm detection by detecting all units in parallel.To address the complex implementation process and time-consuming problem of track tracking,this article proposes using a CPU and GPU cooperation approach.The CPU manages the track database while the GPU accelerates the track judgment of target points,successfully achieving real-time target tracking.