Suspicious Activity Detection

Download Report

Transcript Suspicious Activity Detection

Student: Dane Brown 2713985 Supervisor : James Connan and Mehrdad Ghaziasgar

OVERVIEW  INTRODUCTION  DESIGN DECISIONS  IMPLEMENTATION  PROJECT PLAN  DEMO

INTRODUCTION  Extremely high crime rate in South Africa  Car break-in rate was 16000 in 2009  18 times the rate of USA  Carjacking is the most common crime in South Africa  Costing tax payers billions of rands!

INTRODUCTION cont.

Carjackings 2006-2009

16500 16000 15500 15000 14500 14000 13500 13000 2006 2007 2008 2009

INTRODUCTION cont.

 CCTV cameras  Human monitored  Current solution ineffective  Continued high break-in rate

INTRODUCTION cont.

 Pioneered revolutionary system  Uses computer vision techniques  Automatically detects suspicious activity from a video feed  Detection happens in real-time

INTRODUCTION cont.

 Pioneered revolutionary system

DESIGN DECISIONS  Classification methods  Machine learning such as Haar-like features with Adaboost  Generally training 2000+ sample frames  Why not a classification method?

 Trade-off between speed, complexity and accuracy  There are simpler and more robust ways to differentiate suspicious and normal behaviour.

IMPLEMENTATION  Original frame in RGB colour

IMPLEMENTATION cont.

 Gray Scale and Frame differencing

IMPLEMENTATION cont.

 Motion History Image (MHI)

IMPLEMENTATION cont.

 Blob and movement detection (using MHI)

IMPLEMENTATION cont.

 Blob and movement detection

IMPLEMENTATION cont.

 Blob and movement detection

IMPLEMENTATION cont.

 System determines normal activity  Park car

IMPLEMENTATION cont.

 System determines normal activity  Park car

IMPLEMENTATION cont.

 System determines normal activity  Get out

IMPLEMENTATION cont.

 System determines normal activity  Walk away

IMPLEMENTATION cont.

 System determines normal activity  Walk away

IMPLEMENTATION cont.

 System determines normal activity  Get back in

IMPLEMENTATION cont.

 System determines normal activity  Drive away

IMPLEMENTATION cont.

 System determines normal activity  Drive away

IMPLEMENTATION cont.

 System determines suspicious activity  Loitering next to a vehicle is suspicious

IMPLEMENTATION cont.

 System determines suspicious activity  Loitering next to a vehicle is suspicious

IMPLEMENTATION cont.

 System determines suspicious activity  Loitering next to a vehicle is suspicious

IMPLEMENTATION cont.

 System determines suspicious activity  Loitering next to a vehicle is suspicious

IMPLEMENTATION cont.

 System determines suspicious activity  Loitering next to a vehicle is suspicious

IMPLEMENTATION cont.

 System determines suspicious activity  Loitering next to a vehicle is suspicious

IMPLEMENTATION cont.

 System determines other suspicious activity  Parking, but not leaving the vehicle

IMPLEMENTATION cont.

 System determines other suspicious activity  Accelerating too fast

IMPLEMENTATION cont.

 Suspicious activity detected!

DEMO  1. Normal activity - typical drive away  2. Suspicious - two men loitering  3. Suspicious - Stationary  4. Suspicious - Acceleration

REFERENCES  Davis, J. W. (2005).

Motion History Image

. Retrieved 2010, from The Ohia State University.

 Green, B. (2002).

Histogram, Thresholding and Image Centroid Tutorial

. Retrieved 2010, from Drexel University site.

Trip Atlas

. (2010). Retrieved from Carjacking: http://tripatlas.com/Carjacking#South%20Africa 

Hijacking

. (2010). Retrieved from Arrive Alive: http://www.arrivealive.co.za/pages.aspx?i=2364