Virtual perspective views for real-time people detection using an omnidirectional camera

Considering the growing demand for public security systems the lack of availability of automated surveillance systems is a focal problem. Practical systems have to provide key features like: robust people localization, object detection and event decision, real-time processing and low overall system cost. In this paper we present a real-time indoor people tracking and surveillance system based on a single omnidirectional camera. The proposed approach utilizes linear-time algorithms in conjunction with a hardware-oriented implementation. The software was realized for an embedded microprocessor platform which acts as a smart sensor. The presented system aims at real-world scenarios and targets applications in surveillance, security, consumer electronics, home automation and ambient assisted living.