AFIS Candidate List

[Peter Griffin]

We are in the process of getting an NEC AFIS system in the next few months and I was wondering how people are typically running their searches and how many candidates they are returning so we can effectively create an AFIS policy for our Analytical Method.
We are presently looking at running (per latent):

The Standard Algorithm Rolled
The Standard Algorithm Slapped
ELMA A Algorithm Rolled
ELMA A Algorithm Slapped

Then depending on the crime or location running:

IAFIS (top finger choices)
IAFIS (remaining fingers)
ESSO

The default candidate list is set to 30 which means at a minimum we will have to verify 120 candidates per latent (assuming there was not hit). And up to 210 candidates if the search is continued to IAFIS and ESSO.
After reading the Illinois State Police’s ELMA evaluation we are making ELMA B optional and are looking at dropping our candidate list to 10 for most crimes and increasing that list for major crimes. I have heard that 98% of hits are in the top 10 if there are 16 points charted and it seems that most of the hits from ISP’s study were within that range.
I would like to have a balance between efficiency (low candidate list) and accuracy (not missing hits because of a low candidate list).
So if anyone wants to pipe in with how they are doing things and why (regardless of what AFIS system you are using) as well as where most of their hits are falling in the candidate list I would be grateful.
Thanks
PG

[Bill Schade]

Ok, I'll play

Local Sheriff's Office with an in house AFIS system, a connection to the state database, and ULW connection to FBI
We search Palms and fingers

8 examiners conducting latent searches

"hit" is person per case.
ex. 1 burglary with 2 subjects identified by AFIS = 2
1 burglary, 4 fingers of 1 subject identified = 1
1 burglary, identification of an officer or other "eliminations" =1

Calandar year 2009 1/1/09 to 9/29/2009

937 Hits of which 153 were "reverse hits" leaves 784 Latent vs database hits


Of the 784 hits

668 were #1 respondant
15 were #2
8 were #3
4 were #4
4 were #5

all the way down to hits in the 18th position which is the number of candidates returned on local searches

Position on the list is NOT dependant on number of minutia plotted.

I have data going back 15 years. e-mail or call and we can chat

afisboss@aol.com
727 464-6411

Bill

[mgirard]

This is what my agency is currently doing on our NEC AFIS system, but because we are part of a regional AFIS we also search surrounding States regularly.

1. In-state Standard Algorithm rolled
2. In-state Standard Algorithm slap
3. Out of state Standard Algorithm rolled
4. Out of state Standard Algorithm slap
5. In-state ELMA A rolled (if 15 minutiae or less, minutiae all in same quadrant, or latent distortion)
6. In-state ELMA B rolled (if 15 minutiae or less, minutiae all in same quadrant, or latent distortion)
7. ESSO (there are roughly 10 additional databases we can search) and IAFIS searches depending on crime type

We are currently requesting between 30-36 candidates for every search. The number 36 came about because we were noticing atleast 20% of the candidates being returned did not have the pattern type we were requesting. So potentially we are looking at over 200 images for each latent we search. Most of the latents I search have between 7-12 clear minutiae plotted. Without referencing specific paperwork I think about 3/4 of my hits are in the top 3 candidates. The remaining have ranged upward with one at 29 and recently one at 24.

[Charles Parker]

OK---we have access to a NEC system but here comes the dumb question.

What is ELMA A and ELMA B???

What is ESSO ?????

[Amy Hart]

ELMA is Enhance Latent Matching Algorithm, and there are two versions (A and B). They search the print a little bit differently.

I don't know what ESSO is...

[jpadilla5]

OK---we have access to a NEC system but here comes the dumb question.

What is ELMA A and ELMA B???

What is ESSO ?????

That's how I feel.

All I know is that we run both Cogent and NEC systems. We initially search our local database, then the county, then DOJ. We get the top 15 candidates returned on each entry. The 'hits' are normally in the #1 position, but we have had them as low as the 8th position (that I have personally seen).

[Peter Griffin]

Amy is correct. ELMA is Enhance Latent Matching Algorithm and it is different than the standard algorithm in that additional relationships are considered between minutia. They are most effectively used when there is distortion, minutia are in a single quadrant, or there is a low minutia count (typically less than 15).
ESSO stands for External Search System Other and I believe it is an agreement with WIN (Western Identification Network) and participating ESSO states. It allows us to search other local databases for crimes against persons or if there is a connection or association to that specific region.
PG

[Charles Parker]

OK I am back in.

We have a state NEC system that is 10-Prints and Palm Prints. It is not connected to any other outside state agency so we have no ESSO. The state used to have connectivity to WIN but that went away several years ago. You would have to ask the state the When and Why.

The NEC system is AFIS 21 GWS. To my knowledge there is no ELMA A or ELMA B or a choice for either. I have gone through their books and there is no mention of it. Of the 6 LPE here that do work on the NEC they have no knowledge of ELMA A or B.

We have a Cogent local system that has two Sheriffis Departments feeding 10-Prints and Palm Print into it and an additional 3 Police Agencies running inquiries in it. The data base consists of about 390,000 10-Prints sets and 350,000 Palm Print sets.

We have 2 AFIS Technicians who run the bulk of the latent prints in Cogent. They run about 75 images per day that represent about 15 cases. There are 7 LPE who can run additional latent prints on the cases assigned and who will make a decision as to whether to run it in the State or IAFIS.

Only about 1/4 of the cases run in the local system are run in the state.

For the local system we look at what it puts out for Direct Entry which is set at 20 I believe. For the State most of the time it is 10, but that can be changed to 20 depending upon the judgement of the LPE. For Palm Prints with the state I set it at 4 .

Between the two AFIS Technicians and the 7 LPE we are running about 20,000 latent searches a year representing about 4,000 cases (this is a conservative number---rough estimate). The hit rate which includes Detective Requests is around 20%.

Best I can do from a poor old country bumpkin, knuckle dragging uneducated, unkept, old fashioned latent print examiner who relies upon wishful thinking for my conclusions.