二台のカメラを用いたOpenPoseの骨格抽出による移動ロボットの3次元人物追跡

抄録

In this paper, a new stereo vision framework that combines the processing results of two cameras and obtains only the three-dimensional (3D) information necessary for each task is used. Under this framework, in a complex environment where there are multiple people, we aim to realize 3D tracking of a specific person by performing person detection using OpenPose. The human region image is divided into upper body and lower body from skeleton information extracted by OpenPose. A specific person is tracked by identifying the person based on the color of the clothes. In this paper, we propose a tracking method and the effectiveness of the tracking method is verified quantitatively by experiments.