[fprint] Great job!

[Mark Vytlacil]

On Tuesday 20 November 2007, you wrote:
> Mark Vytlacil wrote:
> > I had just about given up hope that you would get past the export hurdle.
> > It works great! I am using the microsoft reader. I am still getting used
> > to putting my finger on the reader in a consistant manner. It helps to
> > press a little. A configuration option for matching threshold might be
> > nice; maybe in /etc/fprint. I put 3 copies of the fprint line in my pam
> > file so that it gives me 3 chances to match.
>
> I feel you should be getting a better experience from this scanner. It
> works very well for me, but maybe I've just become an experienced
> scanner. Anyway, a few tips:
>
> Attempt to put your finger on the sensor fast -- don't put it lightly
> and then improve the pressure, try and put a decent of pressure on
> immediately (not too much though). Although the software will pause
> until you actually remove your finger from the sensor, it will only scan
> your finger once *immediately* when it detects the initial pressure, and
> it will not resample. (in future we'll improve the software to take a
> few samples of the finger and use the best one)
>
> Clean the sensor with sellotape or blu-tak and then get a good
> enrollment image. Use fprint_demo to enroll as it gives you visual
> indication of what the scan looks like and also the enhanced/binarized
> version.
>
> The only time I've had a false rejection from this sensor/driver is when
> I have accidentally rotated my finger to a fairly significant degree.
> This has only happened once or twice during all my testing.
>
> Thanks for the feedback,
> Daniel

Thanks for the tips. At first, I was not using enough pressure to get a good 
complete image. Then, it seemed that I was approaching the scanner at various 
angles, sometimes getting an image of the very tip and missing the whorl. Now 
I am trying to hold my finger more parallel to the scanner as I press. This 
seems to give me a more consistent result. Nonetheless, I still get some 
rejections and the images in fprint_demo look surprisingly good. The number 
of minutiae detected varies between less than 40 to 70 or so.

While writing this, I have been experimenting some more with fprint_demo. I 
notice that there are relatively few minutiae near the tip of my finger and 
quite a few near the back edge of the image. I think that if I miss just a 
little of that further back area, it might change the match quite a bit.