At 03:16 PM 4/14/96 -0400, Jean-Francois Avon (JFA Technologies, QC, Canada) wrote:

...

Incidentally, this simplicity shows the flaw in using this kind of system as an identifier: Since people's faces are usually visible, and can be photographed in the near-IR surreptitiously, it isn't clear how to prevent faking a face which appears to have the same IR signature and pattern.

I remember in a booklet from Kodak on their Ektachrome IR film, there was a picture of a forearm where all the veins were made clearly visible. This film is near infrared (if I remember, the red color on the film corresponds to around 1100 nm).

1100 sounds pretty far into the IR spectrum for silver-halide film to pick up, but I don't know how far they can "push" film to do this. Silicon CCD image pickups peak at somewhere around 900 nm, but they can probably handle 1100 nm at a reduced sensitivity.

Veins and artery identification might be possible, maybe, since fingerprint identification is possible. A friend of mine developped a quite functionnal algorithm doing just that in the late eighties. OTOH, the blood vessels patterns are probably much more constant, from individual to individual, than fingerprints. Just correct me if I am wrong.

Do you mean "constant" over time? Fingerprints are fairly constant, I assume artery and vein number and location is fairly constant too if major weight gains and losses can be ignored. What I don't know is how unique such blood vessel patterns are, compared to fingerprints. The huge numbers frequently given to show how unique fingerprint patterns are, and thus how reliable fingerprinting techniques are often based on a full set of 10 fingerprints, not just one print.