Computer Vision for Space Robotics: CubeSat Detection and Tracking with OpenCV

Development of robotic manipulators for space operations is crucial to meet demands for small tasks, reducing astronaut risk via autonomous systems. Computer vision techniques provide the space robot with information about the target object for maintenance or repositioning, such as a CubeSat. To fulfill these requirements, this paper presents the development of a monocular vision system with the aim to provide visual detection, tracking and distance estimation of small satellites like CubeSats. The system, built using OpenCV, is designed to identify, track, and estimate the distance of a 1 U CubeSat within a controlled test environment simulating spatial conditions. By leveraging a combination of background subtraction, edge detection, and geometric analysis, available in OpenCV, the system effectively isolates the CubeSat from dynamic backgrounds, maintains robust tracking, and provides accurate distance estimations. Extensive testing across various scenarios demonstrated that the system performs particularly well in object tracking. Despite the challenges posed by varying lighting conditions and the nanosatellite's small size at greater distances, the system proved to be reliable, albeit limited. These findings underscore the potential of the system for applications in autonomous navigation and satellite servicing.