Video surveillance system for tracking human activity

Abstract

The objective of this project is to develop a state-of-the-art video surveillance system which is able to categorize different types of human action in complex background environment. The system makes use of multiple interest point detectors to obtain a large number of features, estimates motion vectors using optical flows and SIFT based matching, correct motion of these features by segmentation and homography estimation, then finally train and test data using vocabulary tree.