Automatic Fall Detection in Real Time Video Based Applications

Abstract

In this study, a fall detection system with real time video tracking is implemented to be used where people must be in continuous supervision. Automatic detection of a falling person from a video image is an important problem especially in security and safety applications such as supportive home environments and closed surveilance systems. In this work, first human motion is detected in videos and then tracked. The approach presented here is different from other approaches in the literature. In our work, images grabbed from a continuous video source are stored and if no motion is detected, previously stored still images are evaluted to decide whether the "no motion" condition is a result of a fall.