Grabs and tests of identifying the location of the pupil using a IR web cam. This is work in progress, I

I am using a custom filter based on the following tutorial http://www.codeproject.com/cs/media/Motion_Detection.asp

Since the motion tracking methods produces too many errors - reflection off the eye itself causes the library to detect motion across the eye, I am attempting to location the pupil on a frame by frame basis. Also the pupil can move faster than the software can pick up.

I think I need to add a IR Led to have a constant light source, at the moment the camera is using ambient levels of IR and thus the success is dependant on time of day, weather etc

The approach I am working on will requires an analysis of the points detected to find the area with the highest/density or even circular in shape.

Headset

Headset in use

Eye Capture

Motion Tracking - Threshold Based (inconsistant)

Motion Tracking - Threshold Based (problems with multiple objects)

Motion Tracking - Threshold Based (raw output)

Motion Tracking - Floyd-Steinberg Binarization (need cluster analysis)

Test Video - Google managed to mangle to quality somewhat