Design Of Low-altitude Flying Target Detection System

UAV have important application value in military reconnaissance,industrial production and agricultural plant protection.However,most UAVs are flying without supervision,which seriously threatens public safety and may even lead to hidden dangers to national security.Therefore,it is urgent to develop a low-altitude UAV target detection system for all-weather monitoring of sensitive airspace and discover UAV targets that affect airspace safety.At present,most of the detection systems designed for low-altitude UAV targets use radar or radio detection methods.Such a system is large in size,high in power consumption,and low in detection precision,and it is difficult to meet the application requirements of low altitude and small UAV target detection in China with a large area and complex terrain.As the precision of image detection algorithms for target detection continues to improve,image detection methods have been widely used in anti-UAV systems.This thesis uses the Zynq-7020 platform,combined with the Tiny-yolo algorithm,to design and develop a low-altitude UAV target detection system that integrates an image acquisition module and an image processing module.The results of the test and analysis of the system energy efficiency ratio and mean average precision index show that the low-altitude UAV detection system designed in this thesis has the characteristics of high precision,low cost,high energy efficiency ratio,and easy deployment.The main contents of this thesis are as follows:(1)The design of the detection system architecture and algorithm implementation scheme:By analyzing and comparing the advantages and disadvantages of different architectures of existing image processing systems in terms of power consumption and versatility,the architecture and algorithm implementation scheme of low-altitude UAV detection system are designed.In view of the shortcomings of the single image input interface of the general-purpose processor,the FPGA is used as the processing unit of the image acquisition module,and the image input interfaces commonly used in industrial cameras such as HDMI,Camera Link and GMSL are also designed.The image processing module uses Zynq-7020 as an algorithm acceleration platform to solve the problem of low energy efficiency ratio of GPU platform.The UAV target detection algorithm uses Tiny-yolo algorithm,which improves the detection precision of UAV targets compared with traditional target detection algorithms.(2)The design of the hardware circuit of the detection system:According to the overall design scheme of the detection system,the hardware circuits of the image acquisition mod-ule and image processing module are designed considering the circuit principle,working stability and shape structure.Designed the image acquisition module HDMI,Camera Link and GMSL three image input interfaces and LVDS data transmission channels and other hardware circuits,and designed the image processing module Gigabit Ethernet,SD card and UART and other peripheral interface circuits and HDMI images Output interface circuit.(3)The development of hardware and software of the detection system:The logic design of the FPGA is completed according to the requirements of the image acquisition module,and the functions of three interface input image data acquisition and image data transmission are realized.In response to the needs of UAV target detection,the hidden layer of the Tiny-yolo convolutional neural network was deployed on the Zynq-7020 PL using the Vivado HLS tool,at the same time,the software part of the input layer and output layer of the Zynq-7020 PS Tiny-yolo convolutional neural network was completed.(4)Testing and analysis of the power consumption and detection precision of the detection system:Compared with the GPU platform and the embedded GPU platform,the power consumption test analysis experiment shows that the system designed in this thesis has the advantages of low power consumption,and the energy efficiency is higher than the index 1.97 times.The mean average precision index of the test system indicates that the system has high detection precision for low-altitude UAV targets.The Vivado tool was used to analyze the resource usage of Zynq-7020 and the power consumption of different parts,and proposed the improvement direction of the system.The low-altitude UAV detection system designed in this thesis has high detection precision for UAV targets,and has a higher energy efficiency ratio than GPU platforms and embedded GPU platforms.At the same time,the system has the advantages of small size,low cost,and easy deployment,which has a certain reference value in the field of anti-UAV system.