In order to meet the requirements of unified and concise description of non-cooperative spacecraft for cooperative awareness, a new method for sparse semantic reconstruction is proposed by using the various components and their relative relationships to describe the spacecraft. Firstly, an affine transformation is used to improve the 3Dbox label method, and the results contain relative depth information. Then, eight vertices of the spacecraft body are identified by using the YOLO6D method to establish the target coordinate system, and an attitude estimation method is given. Finally, the Yolact method is used to detect spacecraft components, and the singularity of the target coordinate system is solved by calculating the relationship between each spacecraft components and spacecraft body. The simulation experimental results show that this method can complete the reconstruction and attitude estimation of spacecraft from a single image. The F-value of reconstruction is better than 0.91, and the relative attitude estimation accuracy of the target is better than 4.5°.
A Sparse Semantic Reconstruction Method for Spacecraft for Collaborative Awareness
Lect. Notes Electrical Eng.
Chinese Conference on Swarm Intelligence and Cooperative Control ; 2023 ; Nanjing, China November 24, 2023 - November 27, 2023
Proceedings of 2023 7th Chinese Conference on Swarm Intelligence and Cooperative Control ; Chapter : 56 ; 661-673
2024-06-15
13 pages
Article/Chapter (Book)
Electronic Resource
English