Study On UAV FPGA-based Vision Tracking System

Due to the investment of major colleges and universities and market capital,the rapid development of UAV related technologies has been widely applied in related fields such as aerial photography,aerial reconnaissance and surveying and mapping.At the same time,the demand for intelligent UAVs has gradually increased.In the field of computer vision,because of its large amount of information and high real-time performance,it can provide massive and real-time external information for the UAV's intelligent flight.However,the current visual tracking system is mainly based on GPU or DSP,and the computational cost is high and the power consumption is high.Therefore,the research on low overhead and low power UAV vision tracking system has far-reaching significance for the acceleration of the UAV intelligent process.Based on the above background,based on the analysis of functional requirements of UAV,the research on UAV visual tracking system and the design of related circuits are carried out.In this paper,the common target recognition and tracking methods are studied and analyzed,and the Cam Shift algorithm and Kalman algorithm are analyzed emphatically.Based on the classic Cam Shift tracking algorithm,aiming at the characteristics of digital images,ROI extraction is introduced and the contribution factor distribution is introduced.The improved Cam Shift algorithm is optimized by using the classifier algorithm.The simulation results show that the improved Cam Shift algorithm has the advantages of low computational complexity,high speed and good performance.Based on the demand analysis of UAV and the research of visual tracking system,a dualmode vision tracking system based on improved Cam Shift algorithm and Kalman algorithm is designed.In order to facilitate the intuitive visual experience of the user,the visual tracking system uses the tracking result to update the original data of the video to display the visual tracking object and the tracking area directly on the video display.The simulation results show that the tracking algorithm of visual tracking system reduces the computational complexity,improves the data processing speed and is robust.According to the system's video caching needs and the actual hardware resources,the use of pingpong operation to optimize the design of storage solutions.For the visual tracking system model and storage scheme for the design and implementation of circuit architecture,and set up the entire hardware system for system debugging and function implementation.This topic uses Spartan6 series chips as the core of the AX845 FPGA development board,OV5640 camera module and LCD display to build a hardware system for test validation.One OV5640 camera module supports 1080 P @ 30 FPS video output,UAV to meet the demand for video clarity and frame rate.According to the functional test analysis of the hardware system,this system real-time tracking,low cost and low power consumption in the environment of moving camera and moving target.Test results show that the UAV visual tracking system is feasible and reliable.