面向卫星在轨实时处理的遥感视频动目标提取算法

针对现有动目标检测算法应用于卫星视频存在较多伪运动误检且难以在轨实时运行,同时短程跟踪算法难以寻回丢失目标的问题,提出一种卫星在轨实时提取运动目标算法.面向运动区域设计图像分类算法以优化运动检测结果,准确筛选动目标;用短程跟踪代替逐帧检测,以降低整体算法复杂度,并设计多特征融合与时空约束的重识别机制关联短程轨迹,应对跟...

MOVING OBJECTS EXTRACTION FROM REMOTE SENSING VIDEOS FOR SATELLITE ON-ORBIT REAL-TIME PROCESSING

Applying existing moving object detection algorithm to satellite video,there are many pseudo-motion false detection and it is difficult to run in orbit in real time.And short range tracking algorithm is hard to find the lost target.Aiming at the above problems,this paper proposes an algorithm for real-time extraction of moving objects from on-orbit satellite.We designed an image classification algorithm for the moving regions,so as to optimize the motion detection results and extract moving objects accurately.Short-term tracking was used instead of frame-by-frame detection to reduce the overall algorithm complexity.We devised a recognition mechanism of multi feature fusion and spatio-temporal constraints to associate short-term track trajectories in case of tracking loss.An on-orbit intelligent real-time processing system was put forward.By using the real satellite video data,the experiment was carried out on embedded simulation platform.The results show that the accuracy and real-time performance of the proposed algorithm is obviously improved compared with the present methods.