1P1-E09 Single-View 3D再構築に基づく頑健な自己位置推定(3次元計測/センサフュージョン)

抄録

The problem of memory-based robot self-localization is approached from a novel 3D reconstruction technique in the paper. A single-view 3D reconstruction technique employing on the "Manhattan world" assumption is introduced for synthesizing views at various possible viewpoints. Experiments are conducted on indoor mobile robots with an omni-directional vision sensor.