Research On The Parallel Processing Technology Of Multi-mode Radar Signal Processing Algorithm Based On GPU

The basic task of radar is to detect the targets of interest,measure the parameters of the distance,speed,position and so on.In order to adapt to the complex external electromagnetic environment and the changeful nature environment,and to complete the different tasks of perceptual,multi-mode radar with a variety of work modes has been an inevitable trend with the development of the modern radar.With the increasing demand for radar application,whitch requires the radar system detect and identify the target accurately,and can quickly complete the imaging with high resolution.Thus,the radar system must deal with large amounts of data within quite a short time,and there is a higher demand for the accuracy and real-time performance of data processing.The traditional processor has been unable to meet the demand for the real-time processing of radar system.CUDA enables the GPU to realize the general-purpose computing conveniently and efficiently.It provides a new solution for the real-time processing of radar,and it is a hotspot to use GPU as the processor in radar system.This paper studies on the parallel implementation of the processing of multi-mode radar based on GPU,of whitch the modes include the SAR mode,PD mode,CW mode,simple pulse mode and so on,and what the paper studies on is from the actual project.In the implementation,CPU deals with the resource scheduling and the logical processing,while the GPU aims at the parallel processing of signal.The content of this paper follows:The kernels and the tasks are processed concurrent using the multi-streams and multi-GPUs: the multi-channel data is processed concurrently,whitch makes a good use of the GPU computing resources;and multi-tasks are dealt with synchronously by using the dual-GPU feature of K80.Comparing to the traditional method using GPU,the solution in this paper improves the efficiency of the algorithm greatly.In the implementation of fine-grained,I optimize the code by decreasing the access to global memory and increasing the parallelism of algorithm,and using the arithmetic instruction reasonably.In addition,the optimization scheme for different processing module is given.At last,I realize the parallel processing using the emulational data from Matlab,and compare the result with Matlab.In this way,I proof the feasibility,accuracy,and real-time performance of the parallel scheme.The experimental results show that the GPU parallel scheme in this paper improves the processing ability of multi-mode radar system to a large extent.The parallel algorithm achieves by 50% in sar imaging and in the measurement of speed and the distance compared with the traditional implementation using GPU,and the parallel scheme can meet the real-time requirements in practical application.