无控制点场景下基于多运动目标的卫星视频稳像算法
Satellite video stabilization algorithm based on multiple moving targets in no-control-point scenarios
李睿瑄 1李峰 1辛蕾 1杨雪 1张南1
作者信息
- 1. 钱学森空间技术实验室,北京 100094
- 折叠
摘要
为了解决无控制点场景下现有卫星视频稳像算法失效的问题,提出了一种基于多运动目标的卫星视频稳像算法,该算法构建了目标检测、轨迹平滑和范数优化的框架,将多目标卡尔曼滤波、rlowess轨迹平滑与L1-L2范数优化相结合,最终实现卫星视频稳像.利用海上多目标舰船观测数据集开展实验,结果表明基于多运动目标的卫星视频稳像算法在X、Y方向上的稳像误差均不超过0.3个像素,验证了该算法的有效性.本研究填补了无控制点场景下卫星视频稳像的技术空白,具有重要的工程应用价值.
Abstract
To address the failure of existing satellite video stabilization algorithms in no-control-point scenarios,a satellite video stabilization algorithm based on multiple moving targets is proposed.This algorithm constructs a framework that integrates target detection,trajectory smoothing,and norm optimization by combining multi-target Kalman filtering,rlowess trajectory smoothing,and L1-L2 norm optimization,ultimately achieving satellite video stabilization.Experiments were conducted using a maritime multi-target ship observation dataset.The results show that the stabilization error of the proposed algorithm can be controlled within 0.3 pixels in both the X and Y directions,and validate the effectiveness of the algorithm.This study fills the technical gap in satellite video stabilization in no-control-point scenarios and has broad and significant engineering application value.
关键词
无控制点场景/卫星视频稳像/多运动目标/亚像元级/卡尔曼滤波/rlowess平滑/L1-L2范数优化Key words
no-control-point scenarios/satellite video stabilization/multiple moving targets/sub-pixel level/Kalman filtering/rlowess smoothing/L1-L2 norm optimization引用本文复制引用
出版年
2024
NETL
NSTL
万方数据