GYRO and AFIS

[GrayMatter]

I've been teaching an analysis workshop with Allison Loll, who works at Phoenix PD, for 5+ years now. The one thing that we ALWAYS hear from our students is that analysis markings are time consuming and isn't clear they are always needed. While we agree on both counts (time and necessity), we still find value in teaching how to do a thorough analysis when it is warranted. One thing I've noted is that no one complains about encoding features to launch AFIS searches. Examiners can't effectively resolve cases without AFIS searches. And we do a lot of work in AFIS to supplement our case records. AFIS could be an excellent documentation component. ENTER a new class- RS&A is offering 'Incorporating Technology into the ACE-V Process'. Currently, scheduled for January 2019. Hosted at Idemia in California so there definitely will be AFIS technology. My hope is that AFIS will be more seamlessly integrated to provide efficiency particularly for analysis.

http://www.ronsmithandassociates.com/tr ... hedule.php

[Steve Everist]

One thing I've noted is that no one complains about encoding features to launch AFIS searches.

Could this be because, until recently, there hasn't been an alternative to encoding of features to conduct an AFIS search? I would guess that, given an alternative, the answer to the question, "If you could do an AFIS search without encoding features, and get the same results (or better) than by encoding features, would you prefer that?" would be pretty unanimous in not encoding.

We're currently still using the MorphoTrak system, with a combination of auto-encoding and examiner-encoding. However, with NGI using MorphoBiz and being one of our search options, I love the simplicity of doing an Image Search. The Image Search requires no encoding, and we've been happy with the results of the NGI searches and definitely prefer the efficiency of that process. We're currently moving to MorphoBiz, and I'm looking forward to the ability to do Image Search as opposed to having to encode each feature for a search. I'll be happy to say goodbye to manual encoding, if the results will be comparable. I would assume Idemia would support the idea that their Image Search function isn't inferior to the old way.

[Bill Schade]

Experience here was that Auto Encoding is equal to operator encoding in most cases. Even terrible latents "hit" when auto encoded. I never thought I would see the day.

The question for me going forward:

Is Auto encoding more "consistent" than examiners. And if it is, do features detected by the computer have more "weight" than a minutia point seen in a smudge that is marked by one examiner and not others. Certainly can't accues a computer program of being biased against a particular subject in a database.

Going forward with technology these questions will be more relevent to your expertise than "Can I find it"

[Boyd Baumgartner]

Experience here was that Auto Encoding is equal to operator encoding in most cases.

It depends on what you're putting through AFIS, and if that's something that's standardized in the SOP, I suppose. I have always been fascinated by the Quality Metrics in ULW and seeing what the distribution of scores to Quality Metrics is (since you don't enter hit/no hit decisions). We were going to track it for a while and one examiner did for at least 20 IDs.

The range was 52 on the low end and 99 on the high end. The distribution was QMs in the: 50s = 4, 60s = 4, 70s = 3, 80s = 3, 90s = 6.

I would venture to say that the more complexity (indicated by lower QM) you introduced, the less auto encoding would be equal to examiner encoding.

[NRivera]

Experience here was that Auto Encoding is equal to operator encoding in most cases. Even terrible latents "hit" when auto encoded. I never thought I would see the day.

The question for me going forward:

Is Auto encoding more "consistent" than examiners. And if it is, do features detected by the computer have more "weight" than a minutia point seen in a smudge that is marked by one examiner and not others. Certainly can't accues a computer program of being biased against a particular subject in a database.

Going forward with technology these questions will be more relevent to your expertise than "Can I find it"

I think you are very correct in that last point Bill. I don't know that I would give more "weight" to a computer-selected characteristic. While the systems have evolved greatly over time, I still see auto-encoded minutiae that aren't really minutiae, but the computer "sees" them that way. If the purpose of the marking is to search a reference database that has also been auto-encoded by the same system, that marking does hold more weight in terms of consistency of feature annotation for that specific system, but not necessarily for the human examiner coming behind it. I guess that's why we make the big bucks. It might be pretty worthless if using that same marking to search in another system that may have a different auto-encoding algorithm (that perhaps doesn't mark as many ghost minutiae in the knowns).