Research On Moving Target Detection And Tracking For Satellite Video

In recent years,with the development of LEO(Low Earth Orbit)satellite,the video satellite with the ability of continuous gaze observation has attracted more and more attention of experts and scholars,and the extraction of moving targets from satellite videos has become one of the hot spots of scholars.Since the satellite platforms cannot provide enough space and energy for the large-scale(size)optical equipment when working in the space,the image acquisition ability of the optical equipment loaded on the video satellite is limited,which makes the target in the video satellite unable to provide enough moving target characteristics.In addition,the satellite needs to adjust its position and observation angle at any time to perform different observation tasks,that will generate a large number of "pseudo targets",the ground targets and the video satellite move relative to each other.Therefore,traditional moving target detection algorithms commonly have the problems of low accuracy and low recall rate when they were used in satellite videos.because tracking errors or tracking loss often occur.Based on the above problems mentioned,this study mainly carries out the flow works:1.A new method named Dave(Deviation limitation Adoptive constraint enhancement detection method for Vi BE)base on Vi BE(Visual Background Extractor)is proposed for moving target detection from satellite video sequence to deal with the low precision rate and recall rate of traditional algorithms for satellite video.Firstly,Frames can be process by Dave with contrast enhancement method in advance which can improve the visibility difference between the targets and the background.Secondly,a remember matrix is proposed to remove the noise and ghost that were caused in the firstly step,such that improve the precision level.Finally,morphology methods are used for removing small noise and enclosing the target.A series of contrast experiments were conducted to validate the advantages of Dave method.At the same time,we also discussed shortcoming of the Dave method.2.A new method named Hovcut(High Order Vector Correlated Update Table)is proposed for moving target tracking to improve the precision level.Every moving object is constructed as a vector by Hovcut and the information of the vector is stored in an adjacency table,and the specific position of moving object is obtained by looking up the table.The position of moving object is calculated by using the motion vector.The moving object is roughly screened by cluster size.Then the cosine similarity and histogram distribution similarity are used to match to get the most After that,the moving target is tracked accurately.Then,the advantages and disadvantages of the algorithm are analyzed by experimental verification and comparison.The moving target detection and tracking algorithms proposed in this study improves the problem of low recall rate of traditional algorithm,reduces target tracking loss and error,and paving solid way for following moving target information extraction from satellite video.The algorithm proposed in this study will be of great significance to the construction of intelligent transportation,Military and other fields.