Design and Implementation of a Real-Time Intelligent Video Surveillance System

Abstract

This thesis presents and proposes several techniques for developing real time intelligent video surveillance systems ranging from simple image processing to the most sophisticated machine learning methods. The aim is to detect and recognize an abnormal situation in a video; in order to achieve that this work is split into three main parts. The first part consists of benchmarking different segmentations and motion detection methods. The second one consists of implementing a real time system for action recognition on hardware circuits, and the last part on developing high efficient system for recognizing fall of older people by proposing a new algorithm for extracting features. The two originalities of this work are the development of this algorithm and the elaboration of a complete benchmark for motion detection methods.