Pupil Detector Hardware

You have reached the project log for our pupil detector hardware.
This log contains the result of a ten week capstone project
completed at the University of California, Santa Cruz in Winter 2009.

We have released our project files.
Free to copy, distribute, display, and to make derivative works.
Please attribute :
John Robertson (jrobot AT gmail DOT com)
Nghia Nguyen (roynguyen84 AT gmail DOT com)

Download All Hardware Schematics, Layouts, Gerber, Etc

Download Project Report

The first week of our quarter has passed and we are proud to present our IR LED driver PCB

It took a few attempts to get the LPKF tweaked to our liking

Pretty nice results, though I should size the through-hole pads a bit bigger

Roy wasted no time to start assembling the board

Next week we hope to have the MT9P031 camera module fabricated and assembled

Another Week!

Roy has brought up our IR driver boards

http://www.vimeo.com/2977760

http://www.vimeo.com/2977741

John has worked with the Santa Cruz camera module to develop a networked camera (the input and output of our image processing pipeline)

Results of incorrectly mapping our image data to a ushort_gray color space

Fixed the mapping and experimented with our 120 degree FOV lens

Next week we hope to improve the performance of our camera module and research the fabrication of our other pcbs

First Private Demo

Discovered that the cameras rolling shutter requires us to precisely control our illumination in order to keep the On/Off-Axis lighting to their respective frames.

Developed a new PWM strobe control to support the rolling shutter. Overcame an odd bug.

This may have been caused by some sort of optical feedback loop with the Auto Exposure Control / Auto Gain Control and the strobe. It has been eliminated by modifying the dependencies of the PWM IR signal.

With the lighting sorted, started experimenting with different threshold values (Using an Avalon Slave register hugely reduced the compile times).

Roy has been hard at work on the hardware
http://www.vimeo.com/3465795
http://www.vimeo.com/3465761

We are proud to present the first test images from our system. The client renders the image data (in blue) and overlays a crosshair at the detected pupils centroid. Preliminary testing has shown an approximate hit rate of 90%. We will be releasing our project hardware, logic, and software under the GPL in the next few days to this site.

The system operates on 864×600 images with an temporal resolution of 67 ms (15 fps).

Roy made some diagrams demonstrating the pupils reflections