Vision-based Control of a CCTV Pan-tilt Video Camera

摘要

In this article, a simple and constrained approach to active camera control is illustrated by a real-time machine vision demonstration for human-computer interaction. The application is driven by the manufacturer's data, contrary to usual approaches that select the pan-tilt units according to the output of the algorithms. The available CCTV pan-tilt camera unit can be controlled solely-using the duration time of a motion command. The output of the image feature extraction module is a set of skin-like pixels. Position monitoring is then achieved by pulling back the center of the image frame towards the position of the center of mass calculated over detected feature points.