US Patent 10553026 Dense visual SLAM with probabilistic surfel map

A novel map representation called Probabilistic Surfel Map (PSM) for dense visual SLAM is disclosed. The PSM maintains a globally consistent map with both photometric and geometric uncertainties encoded in order to address inconsistency and sensor noise. A key aspect of the visual SLAM method disclosed herein is the proper modeling and updating of the photometric and geometric uncertainties encoded in the PSM. Strategies for applying the PSM for improving both the front-end pose estimation and the back-end optimization are disclosed. Moreover, the PSM enables generation of a high quality dense point cloud with high accuracy.