The FPGA Design For Vehicle Target Detection In High Resolution SAR Images

With the rapid development of Synthetic Aperture Radar(Synthetic Aperture Radar,SAR)imaging technology,SAR images obtained by imaging contain more and more data information.Accelerating the processing of SAR image data becomes very necessary.Heterogeneous computing has become an important way for people to speed up the processing of massive data.CPU+GPU is the most common heterogeneous computing architecture at present.GPU has strong computing power and it has a unique advantage in the field of parallel computing.However,with the continuous expansion of GPU clusters,the huge power consumption and heat dissipation problems have gradually emerged,which has become an important constraint that restricts GPUs from doing parallel computing.The emergence of Open CL heterogeneous computing standards has enabled developers to develop in languages without the need to learn traditional hardware description languages,which is a big step for FPGA in the field of high performance computing.The heterogeneous computing architecture based on FPGA+CPU has low power consumption and can flexibly compile hardware according to specific needs.And it has good application prospects in some low-power and high-performance fields.The main work of this paper is as follows:1.In this paper,the vehicle detection algorithm for SAR image based on human visual attention mechanism is optimized and accelerated using FPGA+CPU architecture.Firstly,based on the human visual attention mechanism,the vehicle target detection of high resolution SAR images is realized.Then,the Open CL kernel program is designed according to the specific steps of the algorithm.And the kernel function is optimized by using various methods such as memory access optimization,loop unrolling,vectorization,pipeline replication and data transmission optimization based on FPGAbased Open CL.The algorithm is accelerated on the FPGA platform and this design improves the resource utilization of the system.Compared with the performance of the vehicle detection algorithm based on the CPU platform,the FPGA heterogeneous computing platform has faster execution speed and lower energy consumption and has similar detection results as the CPU platform.2.This paper uses Speed Cloud’s SC-VPX system to accelerate the detection of vehicle targets for large-scale high-resolution SAR images.First,the platform environment for the business board is built to execute the Open CL program.Based on the Zero MQ multi-threaded network library,the communication between the master control board and the business board of the SC-VPX system is realized.Then,the task distribution method is used to make multiple business boards work at the same time and the acceleration of the high-resolution SAR image vehicle detection method is realized.Finally,the execution speed of the large size high-resolution SAR image on the DE5-Net board and the execution speed in the SC-VPX system are compared,and the detection of the highresolution SAR image on the SC-VPX system has faster speed.