Relative pose estimation of uncooperative spacecraft using 2D-3D line correspondences.

The relative pose estimation of uncooperative spacecraft is an essential technology in space missions. However, lacking the prior knowledge of uncooperative targets brings great challenges to accurately measure their pose, which hinders their application in space. To overcome these problems, an accurate and robust pose estimation approach for uncooperative spacecraft is proposed based on a time-of-flight camera coupled with a gray-scale camera. We reconstruct the target model using our hierarchical registration strategy by the ToF camera and extract 2D lines from image sequences captured by the gray-scale camera. We attempt to establish 2D-3D line correspondences for pose estimation, which effectively associates the 2D image information with the reconstructed 3D model. Experimental results manifest that, when dealing with a completely unknown model, our method computes its relative position and attitude with significantly high accuracy.