Study On Moving Objects Intelligent Sensing And Tracking Control For Video Satellite

Small video satellite is a new kind of earth observation microsatellite,which uses video imaging system,transmits real-time video data and is controlled interactively with human in the loop.Confronted with huge data updated by video satellites all the time,human's comprehension ability can not manage to sense this large amount of data;human's understanding can not catch up with the update speed of information.In order to fully mine and utilize the video images taken by video satellite,to make the control method with human in the loop more scientific and reliable,and to make the operation simpler and smarter,online sensing and tracking imaging of moving objects are urgently demanded.This thesis systematically studies intelligent sensing of moving objects and tracking control for video satellite,and the main work include:Firstly,space target detection in video satellite image with star image background is studied,and a detection algorithm using motion information is proposed,which can obtain the trajectory of the target.The effect of satellite attitude motion on image is analyzed quantitatively,which can be decomposed into translation and rotation.Firstly,bilateral filter is used to decrease noise.Then a single frame image is segmented using adaptive thresholding.Considering the continuity of target motion and brightness change,adaptive thresholding is based on local image properties and prior information of previous frame's detection and Kalman filter.Then the algorithm uses the correlation of object motion in multi-frame and satellite attitude motion information to detect the object.Experimental results with video image from Tiantuo-2 satellite show that this algorithm provides a good way of space target detection.Secondly,an intelligent sensing algorithm of moving object based on deep learning is proposed,which can detect the moving object in video satellite image and identify its classification in real time.The algorithm solves the problem preliminarily that the object scale is small,the training samples are few and the algorithm real-time requirements are high.The small-scale characteristics and motion characteristics in the time domain of remote sensing video moving objects are analyzed.Firstly,background subtraction algorithm of modified adaptive Gauss mixture model is used to generate region proposals.Then the problem of training deep neural network under few samples is studied,and a transfer learning method with dimension reduction is proposed.A 21-layer residual convolutional neural network is employed to identify the objects in region proposals,and a tree classification method is proposed.Experimental results about video from several video satellites demonstrate the effectiveness and the real-time performance of the algorithm,.Finally,staring tracking measurement method of moving object by video satellite is studied.A tracking control method based on image feedback is proposed,and the adaptive variable coefficient PD controller for the reaction wheels is designed.The genetic algorithm is used to optimize the coefficients,and the autonomous continuous tracking imaging of the object is realized.The simulation of the control method is carried out,with input of image from video satellite.The general mathematical model of solving the object motion information based on the object image is established.The error analysis of the line-of-sight measurement is studied,and a concise and practical analytic formula of the error range of LOS measurement is derived.A method to determine the motion information of the object on the surface of the earth is presented.Simulation results show that Tiantuo-2 satellite can achieve positioning moving objects on the sea with an better accuracy using this method.Aiming at the absence of the attitude data,a scene matching method based on deep learning is proposed to determine the object position for the detailed background image.The method uses pre-trained deep convolution neural network to extract the features,which have good robustness to light,scale and angle difference.This thesis solves the realtime smart sensing problem of moving objects for video satellite.A tracking control method based on image feedback is proposed.And the object motion information is solved based on the object image.The results of this thesis are of great significance to the application of existing video satellite image and the design of next-generation video satellite,and also of great help to disaster relief,emergency monitoring and battlefield situational awareness.