Aiming at the shortcomings of prevalent visual surveillance system

an intelligent surveillance system which can accomplish automatic detection and classification of moving objects by making use of the computer system is presented in this paper. It mainly consists of such modules as cameras

video capuring

change detection

classification

human interface

alarm log and image database etc. Because current image difference techniaue raises shcu problems as refreshing frequency of reference image and binary threshold selection problems after subtraction of pixel brightness

such comcepts as detecting block

extraction function

binary output function are introduced in this paper to resolve these problems. In addition to

there are complex scenes and lots of disturbance in outdoor environments

so some key techniques such as projection histogram of silhouette pattern

accumulation based on space-time information and geometric model based on edge detection are presented to solve the difficulty of human body recognizing and decrease the ratio of missing alarm and false alarm.