Space mission have involved humans to perform complex tasks requiring perception, decision making, and dexterity. Inherent dangers of space travel, complexity of space systems and paramount safety requirements have been driving up the cost of manned missions. Ground based control via teleoperation offers operator safety, employs remote sensors to provide information for human decision making, and uses remote robots to perform tasks. Communication links with low latency, and high bandwidth and availability are essential for teleoperation. Such links are expensive for Earth orbiting spacecraft and are not at all available for planetary systems. One solution is to increase autonomy of remote agents (servicer spacecraft, planetary rovers, and exploration probes) by providing them with abilities to make their own decisions using data from on-board sensors. Multi-dimensional data such as camera images and three-dimensional measurements from rangefinders offer the most complete representations of observed scenes. Vision systems can be used to process and analyse such data facilitating interpretation and autonomous decision making.
Applications of vision systems for Earth orbit focus primarily on autonomous servicing. Vision guided rendezvous of an unmanned servicer spacecraft with a target satellite may start at a distance of several kilometers and conclude with docking under vision system control. Alternatively, the servicer may hoover in proximity of the target and capture the target with a robotic arm controlled in a visual servoing mode. Robotic servicing will rely on vision systems to model workspaces in 3D and to compare them with blueprints, as well as, to guide end-effectors and tools during operations. Planetary applications that require vision technologies may include landing systems that can guide a lander to a safe landing site and allow for precision landing. Surface exploration with rovers requires vision systems to provide data for local and long range navigation, localization and terrain mapping. Virtual presence and mission monitoring require an ability to create photorealistic and three dimensional models of visited sites.
Challenges of computer vision for use on-orbit include the high contrast and high dynamic range of scenes caused by the intense sun illumination of foreground objects and the lack of diffuse lighting of background objects, and the complex shapes of space structures and their featureless and reflective surfaces. Highly unstructured and self-similar planetary surfaces pose challenges for vision systems for use in planetary exploration as these systems will have to not only detect local obstacles but also incrementally create terrain maps that can be used for vehicle localization and future terrain traversals.
MDA Space Missions has been developing a range of space vision technologies, such as the Spaceborne Scanning Lidar that has been operating since April 2005 on the XSS-11 spacecraft, a vision system that will guide a robotic arm to capture a free floating satellite in the Orbital Express space demonstration, and vision systems for servicing of the Hubble Space Telescope. Several on-going research and development projects address development of vision systems for satellite servicing, localization and terrain mapping for planetary rovers, and environment modeling.