2014 年 19 巻 2 号 p. 237-245
We propose a novel knowledge-based method for rough scale estimation in SLAM using only a single camera. The proposed method assumes that there are multiple planer and periodic-textured objects whose category is given a specific spatial frequency instead of the conventional strict registration of markers or 3-D geometry data. As cues for scale estimation, the method extracts the specific spatial frequency of each object and classifies the object in the corresponding category. A proof-of-concept prototype shows the feasibility of the proposed method through the experiments conducted in a limited scene.