Advanced Spacecraft Detection on Low Power Compute Unit in Space

Obruta, in collaboration with Dalhousie University, aims to advance computer vision techniques for detecting and estimating the pose of spacecraft in space. While object detection in general has made significant strides, existing models often excel at identifying everyday objects but struggle with specialized tasks like spacecraft detection due to the unique and challenging environment of space. Existing pre-trained models require extensive fine-tuning on domain-specific data, which is hard to obtain due to the scarcity of real spacecraft images captured in space. To mitigate these challenges, this research explores two main solutions, namely generating synthetic images using 3D modeling tools and reconstructing 3D cloud representations of spacecraft from sparse photo sequences.

Faculty Supervisor:

Ga Wu

Student:

Partner:

Obruta Space Solutions

Discipline:

Computer science

Sector:

Manufacturing

University:

Dalhousie University

Program: