Pose estimation of daily containers using RGB-D Camera

摘要

Life support robots for assisting handicapped or elderly people in their daily life have attracted increasing attention in recent years. In order to perform the daily tasks such as pick-and-place autonomously, the geometric center of target objects provided by pose estimation is necessary for the robots. In this research, we present an object pose estimation approach of containers in daily life, taking advantages of both color and depth information from an RGB-D camera. The proposed method combined the deep learning based object detection method and 3D point cloud registration algorithm. And pose estimation performance was verified by daily containers of different shape and size in experiments.