Estimating Obstacle Position for Ensuring Safety of Elderly Powered Wheelchair Users

Conventional safety systems for powered wheelchairs, often relying on distance sensors, merely provide alerts about proximate obstacles. This is insufficient for elderly users, who may struggle with rapid perception and agile response, leading to accidents due to the lack of specific obstacle location information. Therefore, this paper proposes a technology to ensure the safety of elderly powered wheelchair users by detecting hazardous obstacles on pathways and providing their precise location information. The proposed method utilizes an RGB-D camera and machine learning for object detection. It then performs a 3D transformation on the detected object’s location within the image, based on the RGBD depth value. This transformation provides the obstacle’s direction and distance relative to the camera’s position, thereby assisting the user’s cognitive perception of the obstacle. Quantitative verification of the proposed technology demonstrates an average position estimation error of 2%.