VISER: View Invariant Shape Estimation for real-time 3D modeling

A high resolution texture mapped 3D model can convey more information than just a 2D map which will be very useful in critical mission planning tasks, security surveillance, exploration tasks, and simulation and training. UC’s proposes VISER to aid indoor navigation and building of 3D database for the ARMY and other DoD agencies leveraging our in-house capabilities, and current projects. Significant features of the proposed VISER for real-time 3D modeling includes: (1) accurate 3D shape estimation of indoor room and objects in the scene using novel in-house Cognitive Object Recognition System (CORS), (2) novel 3D features, improved landmark matching, and realtime relocalization using our in-house algorithms, (3) novel mono-SLAM for robot navigation coupled with object recognition system and next-view-based image capturing for improved performance, (4) efficient hierarchical VRML indoor 3D database which can be ported into different rendering modalities like PC, PDA, or Tablet PC, and maintained with ease and (5) fast and real time rendering techniques for improved user navigation through the 3D environment. The Phase I and Phase I option effort will include developing the proposed VISER technology prototype, building the corresponding key algorithms, evaluating the feasibility under chosen scenarios, and designing the prototype of a system solution.