Near real-time human silhouette and movement detection in indoor environments using fixed cameras

Abstract

The automated human behavior modeling is highly desired in the context of an assistive environment. In this paper, we describe a software video processing and analysis system to assist the near real time detection of human activity. The video data are acquired indoors from fixed cameras in the living environment. The proposed system uses image-processing techniques to segment the human figure from the background, suppress the appearance of its shadow and detect the path and velocity of its motion. Detail performance measurements of the proposed algorithm are given in terms of execution time. Initial results are presented for a small number of video sequences.