The current project tackles in-house safety by communicating domestic incidents such as a person falling or an unusual behaviour. The number of in-house deaths and injuries in the case of elderly, disabled and children could be considerably reduced if these incidents are immediately reported to a facultative or an emergency centre. We believe that the use of multiple video sources will provide us with a powerful, flexible and accurate surveillance/detection system. With this purpose, we envision a distributed smart camera system, based on low-power embedded systems-on-chip targeting image processing and network communication. Our starting point is an existing single smart camera fall detection system developed at our labs within the scope of a research project granted by the strategic fund of the University of Applied Sciences Western Switzerland (HES-SO), which is based on the system-on-chip OMAP1. The implemented fall detection algorithm processes the input video stream by performing, first, a background subtraction, followed by an object segmentation, which serves as input for the tracking of the person and its eventual fall detection.
[read on]
|
