부산대학교 도서관

Fall related injuries are major problems for elderly people leaving alone at home. Most of people who have fallen cannot get up without assistance. Absence of movement of person after fall may cause severe complications regarding health. Considering these problems for elderly we propose a fall detection system for elderly based on computer vision technology. Considering advantages of this technology over wearable based sensor such patients don't need to wear any sensor on body avoiding discomfort due to wearable one. We propose a system in which it is able to detect fall event by detecting features from scene captured by single camera. Features detected are orientation angle, aspect ratio, centre of mass and Hu moment invariants. To reduce computation time in real situation minimum features are used to detect fall. We have also applied IOT (Internet of Things) approach to notify about accident through email with attached screenshot of scene and video recorded after fall. One can also view live streaming of room by entering IP address of raspberry pi on browser if he wishes to see. Camera interfacing and internet connection is done through raspberry pi3 model B which is just like a mini computer with Operating system embedded into its SD card. Overall we propose an automated fall detection system with computer vision and IOT approach.