Study On Terrain Objects Intelligent Sensing And Tracking Control For Small Video Satellite

Video satellite is a new kind of earth observation satellite,compared with traditional remote sensing satellite,the biggest advantage of which is that it can continuously pointing to the target area and achieves real-time observation of hot region.With the development of video satellite,the application requirements have evolved from static observation to real-time observation.Meanwhile,autonomous observation of the hot region and intelligent sensing of potential target have been the imperative requirement of video satellite development.Nowadays,there is little research on the issue of video satellite intelligent sensing,and many technical problems are still to be solved.This thesis systematically studies terrain objects intelligent sensing and tracking control for video satellite,and the main works include:Firstly,terrain background moving target detection with motion information is studied,and a detection algorithm using frame-to-frame registration is proposed.Above all,the SURF feature points of the adjacent frames are extracted,and the SCC of the adjacent frames are calculated to match the frames.Then,the Kalman filter is adapted for proposing the motion information,which solves the problem of the detection algorithm with adjacent frames easily affected by the noise.Experimental results with video images from Tiantuo-2 and Skybox satellites illustrate the effectiveness and the robustness of the algorithm.Secondly,an intelligent sensing algorithm of small scale terrain background object detection with no prior motion information is proposed.First of all,several classical CNN based object detection algorithms proposed in recent years are reviewed.According to the imaging characteristics of video satellite,the two-stage object detection algorithm Faster R-CNN is used as the foundation module,and ResNet-101 is selected as the basic function network of Faster R-CNN.Afterwards,the model is trained and tested on the frames form Google Earth.Experimental results demonstrate the effectiveness of small scale terrain background target detection algorithm.Finally,the problem of intelligent sensing and tracking of Non-cooperative targets under terrain background is studied,and a high stability satellite attitude tracking control based on video feedback is proposed.Initially,a mathematical model based on the video feedback is established under the circumstance of steady staring on the fixed ground point.Then,the coordinate of the target in the image plane is calculated by the terrain background moving target detection algorithm with motion information,and the quaternion error as well as the angular velocity error are derived from the coordinate of the target,which is used as feedbacks to design a video based high stability attitude tracking PD control approach.Finally,the high stability satellite attitude tracking control based on video feedback is simulated with Tiantuo-2 and Skybox satellite imagery.The image processing error,angular velocity measurement error and the interference moment are considered in simulation.The simulation result illustrates that compared with the traditional attitude control method the proposed algorithm not only achieves intelligent sensing and tracking of Non-cooperative targets under terrain background,but also improves the accuracy of attitude tracking control.