Paper about using UV for facial recognition

[Andrea B.]

CITATION:

Samartzidis, T., Siegmund, D., Gödde, M., Damer, N., Braun, A., & Kuijper, A.:

The Dark Side of the Face: Exploring the Ultraviolet Spectrum for Face Biometrics

In: IEEE Computer Society: 2018 International Conference on Biometrics (ICB). - IEEE Computer Society Press, 2018, 8 p., ISBN 978-1-5386-4285-6

 

EXCERPT:

For the UV capturing, a Baumer VCXG-13M camera with a 1/2 CMOS sensor, 1280x1024 pixel resolution and a sensitivity starting at 300nm was used. In order to receive best response of UV-MFP, we filtered the visible light in the captures by using a UV/IR bandpass filter (Schott UG11) which has its peak transmission at 300nm and blocks light with wavelength over 400nm. Furthermore a quartz lens was used to increase the light transmission in that wavelength. For illumination we used two 36W UV-A LPS lamps with a bandwidth between 315nm and 400nm positioned in front left and front right to the subject. In Figure 2 the sensitivity of the camera sensor, the transmission of the used filter and the emission of the used illumination in UV spectrum is shown.

 

FIGURE 2:

 

 

---

I am curious about the filter having a peak at 300 nm and blocking past 400 nm? This doesn't seem like it could be just plain UG11 do you think?

Also, if the lamp has bandwidth between 315-400 nm, then why does the chart show the lamp stopping just before 350 nm?

 

.

[ulf]

I am curious about the filter having a peak at 300 nm and blocking past 400 nm? This doesn't seem like it could be just plain UG11 do you think?

Also, if the lamp has bandwidth between 315-400 nm, then why does the chart show the lamp stopping just before 350 nm?

 

.

 

The text do not match the graphs, but the filter curve is quite similar to an UG11, 1mm.

 

Conveniently their graph stop at 650nm, just before the IR-bump.

 

The light source in the graph seams to be a hand drawn curve intended to look as an UV-LED ca 650nm.

[dabateman]

Andrea,

There is something seriously wrong with this sentence:

"For illumination we used two 36W UV-A LPS lamps with a bandwidth between 315nm and 400nm positioned in front left and front right to the subject. "

 

A low pressure sodium lamp (LPS) has zero UV. That might be a manufacturer I am not aware off.

But its possible they are using some light that ends at 350nm. Like a black compact fluorescent bulb.

 

I was surprised how monochromatic my compact fluorescent black light were when tested here:

https://www.ultravio...rescent-lights/

[Andrea B.]

We can only hope there was no ambient IR in the experimental setting!

[Andrea B.]

Found another paper about UV facial recognition. It looks to me like those poor people in the experiment were UV-flashed with their eyes open? Yikes!!!! This is a PDF link.

Can we match Ultraviolet Face Images against their Visible Counterparts?

 

This group of researchers was using a medical UV imaging system made by Faraghan Medical Camera Systems. The gear is based on a Canon DSLR, has two flashes and is mounted on a frame suitable for facial photographs. I'm not sure what is in the hotshoe. Another flash?

They have a gallery of faces. Some of their subjects, but not all, have their eyes closed.

 

Interesting remark on their website:

Now we have the only portable digital UV camera on the market.

I'm thinking this must be in the context of a medical office where they can roll the setup around from place to place? Yes, that must be it.

 

Here is a good article from their site: http://uvcamera.com/.../Aesthetic2.jpg

[JMC]

I don't even know where to begin with that setup in post #5 Andrea. I followed the link to the website, and found the following statement - "We maximize the Canon’s Digital SLR CMOS sensor to be most sensitive sensor in the UV range on the market!" That's quite a statement.

 

Did you find a transmission spectra of their filter anywhere - they say they are using their own filter?

 

As for the only portable system, what a load of rubbish. Either their understanding is about 10 years out of date, or they haven't bother looking around.

[Andy Perrin]

Or they are simply lying. Advertisers do that.

[Andrea B.]

Capitalism vs. Science: Round #1,879,234,449,876.

[Andrea B.]

A broader point: The Faraghan founder had a good idea. The parts are simple: a converted camera, some flashes with UV-pass filters, a UV-capable lens, a bunch of aluminum tubing and a chin cup. Buy for cheaps, sell for thousands. Anyone of you here could start this business and perhaps make a few dollars/euros/pounds/bitcoins. I would include little eye-cover googles to block the UV flash.

[JMC]

Reminds me of the old snake oil salesmen.....

[Andrea B.]

All advertising is snake oil!! :lol: