INTELLIGENT VIDEO SURVEILLANCE OF HUMAN MOTION: ANOMALY DETECTION

Intelligent video surveillance is a system that can highlight extraction and
video summarization that require recognition of the activities occurring in the video
without any human supervision. Surveillance systems are extremely helpful to guard
or protect you from any dangerous condition. In this project, we propose a system
that can track and detect abnormal behavior in indoor environment. By concentrating
on inside house enviromnent, we want to detect any abnormal behavior between
adult and toddler to avoid abusing to happen. In general, the frameworks of a video
surveillance system include the following stages: background estimator,
segmentation, detection, tracking, behavior understanding and description. We use
training behavior profile to collect the description and generate statistically behavior
to perform anomaly detection later. We begin with modeling the simplest actions
like: stomping, slapping, kicking, pointed sharp or blunt object that do not require
sophisticated modeling. A method to model actions with more complex dynamic are
then discussed. The results of the system manage to track adult figure, toddler figure
and harm object as third subject. With this system, it can bring attention of human
personnel security. For future work, we recommend to continue design methods for
higher level representation of complex activities to do the matching anomaly
detection with real-time video surveillance. We also propose the system to embed
with hardware solution for triggered the matching detection as output.