Design of Ship Target Detection System Based on FPGA

Marine ship target detection needs to process the field data in real time under certain energy consumption circumstances. Traditional methods such as CPU or GPU can hardly meet the requirements of detection accuracy, detection speed, and power consumption at the same time. To address the above problems, this paper proposes a design of a ship target detection system based on the FPGA. The algorithm of this design adopts the improved YOLOv5, and the hardware adopts the Zynq UltraScale+ MPSoC ZCU104 development kit. In hardware development, a custom platform architecture is created by Vivado, Petalinux, and Vitis toolchains. In algorithmic deployment, the ship target detection algorithm is transplanted through the Vitis AI framework. Seaships dataset is used as experimental data. The test results show that the system has an average accuracy of 79.8% and an inference speed of 51.3 fps on the test set. This system reduces required power compared to CPU and GPU solutions.