Research On Space-based Target Extraction And Tracking Algorithms Under Complex Background

Space-based compound axis target tracking system is an important way to achieve high precision target tracking.Compound axis tracking system with appropriate target tracking algorithm can be used in space debris removal and other fields.In this paper,the target tracking algorithm in the space-based compound axis target tracking system is taken as the research object,namely the tracking algorithms under different working conditions are studied.The main contents of the paper include analyzing the general requirements of the tracking algorithm based on the space-based compound axis tracking system and designing the related algorithms.The general algorithm requirements for the space-based compound axis target tracking system are proposed by analsying error mechanism of the tracking system and the target and noise imaging analysis of space-based camera.The algorithms under different working condition,namely the space background aera target,the earth background point target and the earth background aera target,are deisgned,according to the requirements,and the algorithms are verified by some experimentsThe main innovations of the paper are as follows:1)A space background aera target tracking algorithm based on space-based system is proposed,which is based on the Kalman filter,the SURF algorithm and crosscorrelation algorithm and can solve the problems of in-plane rotation,target blur and scaling.The algorithm uses the Kalman filter as prediction segment,the SURF algorithm as correction segment and cross-correlation algorithm as tracking segment to achieve sub-pixel precision and high speed tracking.Through using the SURF feature point's main direction and feature level,the correction segment correct the tracking segment's template.The prediction segment uses Kalman filter to predict target position in the next frame so that to improve tracking speed.The algorithm can handle the target in-plane rotation caused by compound axis system's corase tracking system,short time image blur caused by fine tracking system and the scale change caused by distance between target and camera change.Meanwhile,the algorithm can also handle the brightness change,camera's Gaussian noises and bad points.The final results through simulation image test and model tracking test shows that this algorithm is able to deal with ±125° in-plane rotation and 0.5~2 times scale change,and achieve maxium error less than 0.9 pixel.2)A earth background point source target tracking algorithm based on space-based system is proposed,which is based on the spatial and frequency characteristics of point source target and area noise of the earth background and uses the characteristics to solve the problem.The algorithm uses the frequency filter to find candicate point and spatial domain condition to find the target which achieves high speed and precision and at the same time,the algorithm has good resistance to image noise and camera bad points.Through the simulation image test and real tracking system test,the algorithm is proved to have good performance and no impact on tracking system performance.3)A earth background area target tracking algorithm based on space-based system is proposed,which aims at solving the problem of background clutter and occlusion.The algorithm are based on background-aware correlation filter,and uses histogram of gradient to build background model.The background model is used to detect the target motion and occlusion,combining with the filter response to distinguish target condition,which is used to build the learning set properly.The Object Tracking Benchmark test and real target tracking result shows the algorithm is good at handling target occlusion,deformation and background clutter problems.?The proposed algorithm can solve most scenarios of space-based compound axis tracking system operating conditions,but may not be applicable to some extreme cases.