Visual activity monitoring : assisted child-minding home camera surveillance system

Abstract

The main focus of this project is to develop a robust activity monitoring system which is able track and automatically recognize the activities of a child in an unsupervised environment. Codebook method is used to learn the background for foreground detection and tracking is initialized for every new foreground object detected. Linear Support Vector Machine Model is used for activity recognition of these foreground objects. In addition, a graphical user interface will be provided to enable the user to predefine activity zones. Child wondering beyond these areas will cause the system to trigger a warning signal. User may retrieve earlier surveillance records through this user interface.