The Research On HD Video Electronic Image Stabilization Algorithms Of UAV

In recent years,the UAV has become a focus of research for its flexibledeployment,multitasking and fine cost performance.It has been apllied not only inmany civil fields but also for military.The video surveillance is indispensable toaccomplish the tasks like real-time ground monitoring.The sudden and serverechanges of flight maneuver,the random vibrations are all existed in the flight,and itwill lead to the blurs and unstable phenomenon,the quality of video sequences will beaffected severely.Nowadays,the HD video technology with its high definition,widesurveillance view and other advantages is widely applied.The electronic imagestabilization(EIS) can realize the HD video stabilization with low cost,high precisionand outstanding adaptability and it has a good application prospect in the field ofUAV video surveillance.The paper carries out the research work on EIS,mainly the motionestimation,motion filtering and motion compensation.The gray projection matchingand circle projection matching are combined to estimate the motion vectors betweenframes,which are brought by the translation and rotation motions.According to thecharacterastics of UAV video images,the Sage-Husa self adaption filtering based onKalman filtering is modified for seperating the active motions and random jitters ofthe carrier.The blurs within the frame will be cleared.The multi-vectors averaging ismodified further,when there are noise and obvious changes of brightness,gray andcontrast ratio between frames,the advanced algorithm can still compensate the motionvectors well.Based on the simulation and experiments of the algorithms mentionedabove,the paper transplants the Linux OS to the target board.With the Davinci and HDcodec technologies, the electronic image stabilizing algorithms are optimazed.Then the program is tested on the high speed SOC based on DM6467of TI,so that it can beable to process the UAV HD videos.The experiment results show that the clear and stable video sequences can beaquired afer the processing of the algorithms in this paper,and the results are goodrefferences for the applications of EIS for UAV surveillance videos.