Transcript Slide 1

Student: Ibraheem Frieslaar
Supervisor: Mehrdad Ghaziasgar
OVERVIEW
 INTRODUCTION
 USER INTERFACE SPECIFICATION
 HIGH LEVEL DESIGN
 LOW LEVEL DESIGN
 DEMO
 REFERENCES
 QUESTIONS & ANSWERS
INTRODUCTION
 Eye ball tracker
 Using the webcam to determine eye gaze
location
 Moving the mouse to that location
USER INTERFACE SPECIFICATION
 Runs in the Background
 It’s a Daemon
High Level Design
Input
Webcam
Image
Processing
Move Mouse
Pointer
Low Level Design
Input
Webcam
Haar
Detection
Image
Processing
Locate
Eyes
Determine
left Eye
Move Mouse
Pointer
Locate
Iris
SVM
Low Level Design
• Acquire Image
Input
Webcam
• cvQueryFrame( capture )
Low Level Design
• Haar like features
Haar
Detection
• Detects Eyes
cvHaarDetectObjects
Low Level Design
• Set Haar Detection as
Region of Intrest
Locate
Eyes
cvSetImageROI
• Half the image for left
eye
Determine
left Eye
cvSetImageROI/2
Low Level Design
Locate
Iris
• Hough Circles
• Determine Darkest Pixels
• circles = cvHoughCircles
•
DarkPixels = DarkPixels +
(255-s.val[0])
• MaxDarkPixels = DarkPixels
• cvCircle
• Display Circle
Low Level Design
Right Eye
SVM
• Eye Redrawn
• Pixel values saved
1 1:255 2:255
3:255 4:0 5:0
6:0 7:255 8:255
system("svm-predict.exe Test
trainer.model result")
• Values sent to Support
Vector Machine (SVM)
• SVM Prediction
Low Level Design
Move Mouse
Pointer
SetCursorPos(x,y);
Demo
REFERENCES
 OpenCVWiki. lOpenCVWiki. [Online].
http://opencv.willowgarage.com/wiki/FullOpenCVWiki
 Heiko Drewes, "Eye Gaze Tracking for Human Computer
Interaction," Ludwig-Maximilians-Universität, Munich,
2010.