Gabor Score (Quality Metric) in AFIS Standard

[Boyd Baumgartner]

Does anyone use the Gabor score (Read: Quality Metric) in either ULW or your local AFIS (We have MorphoTrak which has it as well) as part of your AFIS standard? The idea being it's a proxy for the clarity of the print and the resulting minutiae would be more objective. The result being more overlap between the encoding of the latent and the known with a higher chance of AFIS producing the matching candidate.

If you want to nerd out and read what a Gabor Filter is and how it works, go here. spoiler alert: It's an edge detecting filter that the encoder uses to find minutiae/ridge flow.

[Dr. Borracho]

And all this time, I thought Gabor referred to either Zsa Zsa, Eva, or Magda. Silly me.

[josher89]

I really liked the idea of LQM that came out with ULW a few versions ago. At the time I was using it, our state didn't have a connection to IAFIS so we were one of the agencies that had direct latent connectivity to the FBI. When they introduced LQM sometime after version 5.0 (I don't remember which one), I was using it only as a testing mechanism and not as a stat to keep track in our case notes.

I have to say, it did a good job of predicting whether or not the latent would be ID'd if the candidate was in the system. You got pretty 'heat' maps of the latent that became useful for research projects when it was incorporated into PiAnoS.

We were going to begin capturing that information in our notes after we've had the opportunity to use it but sadly, the FBI went away with DLC's in 2017 so our state created a portal to the FBI and we're using that now instead of ULW (since the end of 2016).

[Boyd Baumgartner]

Interesting

Our State system is NEC and you can encode through their system and it can send/receive via the NEC interface, or you can encode in ULW and upload the .lfis or .lffs files directly and you'll receive the .srl files with the NEC system acting as the traffic cop only.


That being said, I'm interested in thoughts about this information from the help file in ULW. Given that we're discussing an articulation document at the moment, the articulation of anything around AFIS (standards for running, how it works, what do scores mean, etc) is severely lacking.

That being said, consider these questions:

• Do you consider low count AFIS Runs to be anything under 10?
• If you run under 10 or whatever you consider 'low count', do you attempt to lower the file penetration (make the pool of candidates smaller) by putting in pattern type, finger position, etc?
• Would a low Quality Metric in combination with a low count either dissuade you from running it or persuade you to alter the file penetration?

.

ULW.png

P.S. Speaking of standards, we need them for humor too. We aren't accountants here... I refuse to acknowledge Dr. Gazpacho's joke as anything but geriatric.

[Dr. Borracho]

P.S. Speaking of standards, we need them for humor too. We aren't accountants here... I refuse to acknowledge Dr. Gazpacho's joke as anything but geriatric.

Second Millennium humor. You wouldn't understand.

[Alan C]

And all this time, I thought Gabor referred to either Zsa Zsa, Eva, or Magda. Silly me.

Kudos for remembering Magda.