パーシステントホモロジーを用いた単純形状物体の位置姿勢推定

抄録

Marker-based estimation is one of the most commonly used methods for estimating the position and orientation of objects. However, there are many problems with markers, and marker-less estimation methods are currently being sought. A common method is to use machine learning to estimate position and posture from distance information and RGB images. However, this method has low accuracy for simple shapes. In this study, we propose a method to simplify position and pose estimation by using persistent homology, a topological theory, to extract geometrical features from distance information and predict the ground plane by machine learning.