6-DOF Pose Estimation for Spacecraft Based on Soft Classification of Euler Angles Optimization

Pose estimation has important applications in on-orbit services such as spacecraft rendezvous, docking, and debris removal. This has become a highly challenging task due to varying lighting, cluttered backgrounds, and occlusions. This paper proposes a method for 6-DOF pose estimation based on soft classification of Euler angles optimization. First, we propose a multi-scale convolutional network with skip connections for better feature extraction. Second, optimization of Euler angles based soft classification is adopted for pose estimation based on the aforementioned network, and its core idea is that the network outputs the similarity of generated orientations and predicted orientation. According to evaluation results, our method achieves high accuracy on the SPPED+ dataset. This paper has also demonstrated the effectiveness of the proposed method through a series of ablation experiments and comparative validation.