Dynamic Scene 3D Reconstruction of an Uncooperative Resident Space Object

Characterization of uncooperative Resident Space Objects (RSOs) plays a crucial role in

On-Orbit Servicing (OOS) and Active Debris Removal (ADR), allowing missions to assess

the geometry and motion properties of the target. Addressing the specific challenge of

reconstructing tumbling targets, this study evaluates the performance of existing state-of-the-art

3D reconstruction algorithms for dynamic scenes (D-NeRF, TiNeuVox, HexPlane), focusing on

their ability to generate physically plausible representations with high fidelity. Specifically, we

address a scene that is both geometrically dynamic, due to the tumbling target, and subject

to dynamic lighting, where directional sunlight casts shifting shadows across the surface. To

support our evaluation, we developed a simulation environment using Isaac Sim to generate

physics-accurate 2D image sequences of tumbling satellites under high-fidelity orbital lighting

conditions. We observe that while HexPlane excels in stationary scenarios (PSNR > 35 dB),

all methods degrade substantially as complexity increases. In fly-around scenarios, D-NeRF

produces blurred outputs, HexPlane exhibits ghosting artifacts, and TiNeuVox introduces foggy

artifacts. Moreover, standard metrics fail to capture these structural inconsistencies due to the

dominance of the black background in the images.