Information acquisition about the scenario might include detecting changes compared to available maps or 3D models of the environment. To determine those changes, newly acquired SfM data and existing 3D models need to be compared. The comparisons requires the alignment of partly matching data and establishes a second fundamental computer vision problem termed Point Cloud Registration.
Finally, information about people and objects that move in the environment need to be gathered: they need to be located, their paths need to be estimated, and their identities need to be determined. These three prominent computer vision problems are addressed by methods for Visual Object Detection, Visual Object Tracking, and Visual Object Recognition.
The mentioned problems are addressed in a range of ongoing research projects such as the Horizon 2020 project CENTAURO on Disaster Response robotics (www.centauro-project.eu), the WASP project on Integrating Perception, Learning and Verification in Interactive Autonomous Systems (http://wasp-sweden.org/research/projects/), and the newly granted SSF project SymbiKBot.
WASP, or rather the industrial main players, focus on three different areas (vision-related adaptation of original goals):
1. Autonomous driving; cars, trucks, mines. Collaboration with Scania and Autoliv
2. Safety, security, rescue and disaster response robotics. Collaboration with SAAB
3. Smart cities, including security and integrity in autonomous camera surveillance. Collaboration with Ericsson