"Automated Surveillance and Alert Generation System"

Authors

E Rama Lakshmi & Hemanth Reddy Loka
Department of Information Technology, Chaitanya Bharathi Institute of Technology, Hyderabad, India.

Abstract

Now-a-days CCTV cameras are present everywhere. A human surveyor monitors the events going on there. There is a lot of burden on the human surveyor, Our idea is to reduce the burden on the human surveyor by automating the process of detecting a disturbance in the crowd. A critical amount of time may be saved by detecting a disturbance in the crowd instantaneously. This critical time we save may be the difference between life and death. This disturbance may be caused by any event such as bomb blast, fire, riots, etc. This paper focuses on the implementation of an algorithm that can detect a disturbance in the crowd. It considers flow-vector magnitudes change over time for a short set of frames to statistically determine whether those short set of frames are violent or non-violent. The challenge is to keep the processing quick and real-time, an alert should be generated within few seconds of the change in crowd behavior. Keywords : Violent Flows, Keras, Optical Flow, Video Processing.