SMARTZOOM
INTRODUCTION

Smart Room at National University of Singapore, deals with the problems of tracking & locating people in a room, detecting events, responding to the events and recording faces of the people who entered the room. Results of the above problems will help in more advanced programs like face recognition algorithms and other intelligent programs.

OBJECTIVES

Smart Zoom is a part of the Smart Room project with following real-time objectives.

  • Detection of human Object, irrespective to his or her ethnic group.
  • Getting the trajectory of the person moved inside the room
  • Controlling the zoom and focus to grab the face of the Person to submit for Facial recognition.

architecture

ALGORITHM

Person presence is confirmed if,
Skin colour is present in Camera #2 AND Object is present in Camera #0
roomsetup

Skin colour is detected using the YCbCr colour space model, approximated to a trapezoid,
317 – 1.67Cb < Cr < 358 – 1.67Cb AND 138 < Cr < 196 – 0.22Cb
ycbcr1

Face is the blob having maximum area with following properties.
Height to Width ratio of bounding rectangle > 1.7 AND Area of the bounding rectangle > 2.5 x Area of the current blog

Tracking is done by finding the intersection of two vertical planes.
tracking

Initially camera #2 is “zoomed in” to get a high resolution picture of the face, as the person enters the room, the lens is “zoomed out” to increase the field of view. Focus is set automatically for both occasions.
zoomfocus

SMARTZOOM IN ACTION

RESULTS

results