MediaPipeを用いた2次元ランドマーク差分に基づく歩行者転倒検知

抄録

In this paper, we propose a fall detection method for elderly people due to illness or accident using camera images and MediaPipe. The method extracts 33 landmarks of a pedestrian and uses the difference ratio of XY coordinates of the landmarks obtained when the pedestrian falls to identify the fall. First, we explain the method for determining the fall state in order to detect a fall. Next, we report on the usefulness and applicability of using the aspect ratio to determine the fall state. Finally, we report on the usefulness and applicability of the method for discriminating between falling and crouching motions.