Model-Based Visual Tracking Of Orbiting Satellites Using Edges
2017 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS)(2017)
摘要
Estimating the pose of orbiting satellites is a key prerequisite for supporting autonomous proximity operations in space. This work presents a monocular 3D tracking algorithm that tracks edges with the aid of an arbitrary 3D mesh model assumed to capture a satellite's shape. The proposed tracker propagates a pose hypothesis between successive frames, using it first to render a depth image and then refining it according to partial matches established between depth and intensity edges. Edge matching relies on fast, local 1D searches along the depth gradient direction. The tracker does not require any preprocessing of the 3D model nor does it make any assumptions regarding its characteristics, as is often the case for other approaches. It is also robust to parts of the tracked satellite being out of view, occluded, shadowed or visually undetected. Experimental results evaluating the accuracy of the tracker and comparing it with established techniques are also included.
更多查看译文
关键词
edge matching,local 1D searches,depth gradient direction,tracked satellite,autonomous proximity operations,monocular 3D tracking algorithm,arbitrary 3D mesh model,orbiting satellites pose estimation,model-based visual tracking
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络