Target tracking using webcams

Abstract

With rapid proliferation of video cameras in public places and ever decreasing price/performance ratio of computing, the ability to identify and track people creates tremendous opportunities for important business and security applications. This report presents a sub-project of the Multiple Camera Indoor Surveillance system. In this project, two webcams are connected to a computer in which C language platform with OpenCV libraries (open source) are installed and knowledge based techniques are used to track and localize a person in a predetermined area.

Using built-in Haar-like feature detection classifiers in OpenCV, we can identify human faces and subtract them out of the background which is normally unchanged and is really not of interest. Haar-like features encode some information about the class (face in this project) to be detected using the existence of oriented contrasts between regions in the image.