Design And Implementation Of Electronic Image Stabilization For UAV Based On FPGA

Unmanned aerial vehicle(UAV)has become a research hotspot in the world because of its flexible,fast,simple structure and low cost.Aerial photography technology is widely used in reconnaissance,search and rescue,panoramic photography,natural environmental protection and other fields.In the course of flight,the mechanical vibration,the influence of wind force and the abrupt change of the motion of the UAV will make the video appear blurred and jump.Electronic image stabilization technology has the advantages of high precision,flexibility and low cost.Therefore,the use of electronic image stabilization technology for UAV aerial video image stabilization has a good application prospect.In this paper,the electronic image stabilization technology of UAV is studied deeply,and an effective method to solve the problem of UAV video sequence jitter is put forward.Based on the study of the basic principle,system model and implementation method of electronic image stabilization algorithm,and combined with UAV's own characteristics and the actual aerial environment,a solution combines of SURF algorithm and adaptive Kalman filter is proposed to stabilize aerial video sequences,the proposed scheme is validated by system simulation and actual experiment,finally,the electronic image stabilization system based on FPGA is relized.In this paper,based on the similarity transformation model and SURF algorithm,it analyzes the common image motion mathematical model,a similar transformation model is developed for the motion of the UAV,this model contains four parameters of motion vector to meet the basic requirements of the image stabilization system.At the same time,the complexity of the model satisfies the real-time requirement of the UAV electronic image stabilization.In the motion estimation,the SURF algorithm is applied.In the design process,the SURF algorithm is modified.The SURF algorithm used the pipeline technology,in the design of non maxima suppression,the original algorithm of serial calculation of structure design for parallel structure,finally,the two layer feature detection of SURF algorithm is implemented and verified.The experimental results show that the SURF algorithm implemented in this paper can accurately match the image features with complex geometric distortion and illumination.The adaptive Kalman filter is designed and implemented in the motion filter.The motion of the UAV is complex,it has not only the fast climbing movement,but also the stable hover.In this paper,the adaptive Kalman filter can change the filter parameters in real time,and achieve the balance between the motion tracking and the motion smoothing.In this paper,an efficient hardware architecture is designed and implemented based on Xilinx Kintex7-325 t FPGA.Based on the analysis of the results and the experimental data,the hardware system can be used to stabilize the 640 × 480 resolution video in real time,and the average PSNR value of the video image is improved by 5dB.