Eye Tracking Project Project Supervisor: Ido Cohen By: Gilad Ambar Eli Yafe Winter 2009 (Tasha) Project B
Background Open CV and our system The system is built from IR camera and 3 LEDs Was designed by ELBIT for eye tracking
OPEN CV OpenCV is a computer vision library originally developed by Intel. (open source) Main purpose real-time image processing Using imaging filter Graphic structures
The Problem What are you looking at right now? Is it in real time? Why not matlab?
Our algorithm in OPEN CV Algorithms Input Output Main Algorithm. Calculating LOS vector Calibrating by off-set vector Current frame LOS vector on screen Calibration Vector Find LEDs and their center of mass Find Pupil Center Pre-calculations and filtering Getting all the blobs* Hough transform algorithm per blob RUNS PARALLEL Select the best ellipse and it’s center *Blob- Binary large object
An output example LOS vector LEDs
The Product Integrated platform to calculate LOS vector Input – stream of images Output – stream of vectors (x,y,z) Parallel work for maximum speed Using designated hardware to speed up the Hough transform Using appointed memory per images for quick use of filtering
The next step GIDEL
The key phrase - Running parallel ! Main program Each blob
Timeline 20/12/09 – Displaying arithmetic operations required. 3/1/10 – Block schematic (Highest resolution. Resources requirements. 10/1/10 – implemented code in fixed point. 11/1/10 – Begin implementation of Hough Transform in VHDL. 12/1/10 – 2/3/10 – TBD. 2/3/10 – Mid project presentation