Can A Surveillance Simulator Help Operators?
The police force in Rotterdam, the second-largest city in the Netherlands, has cut back live operator training and instead uses a program called EyeObserve to train operators. The police say they found it more expensive and time intensive to train operators on the job and now the training with the simulator is mandatory. Has anyone worked with a municipality using something similar or had a request from an end user for something like this. Here are EyeObserve's demo videos:
and
Thanks, Robert. Carlton has contacted them as well. If and when we get feedback, we will share what we learn as well.
That particular claim of having to wait for real incidents to happen when training new operators I found funny... they've got cameras and people already.
Why can't they simulate situations themselves without the need of AI-driven cartoon robots in a game simulator?
I was waiting to see if someone else (maybe involved in the training of surveillance operators) commented on this product first, but since nobody else has, I will: I think it's a friggin joke. :)
I'm sure it's less expensive and time consuming than actually training new operators on live cameras (with real and staged scenarios, etc), but I fail to see how - in essence - playing SimCity: Unruly Crowds Edition would do much to properly train new employees on how to effectively watch real crowds.
VSTEP was developed in part by the dudes who are using it. They have an incentive to say that it's use has actual value.
However, I can't imagine many other municipal surveillance entities rushing over to The Netherlands for on-site demos....
Okay, I'll bite.
Standard operating procedures for a monitoring station is a binder. Lots of situations there that need testing. Operators need to follow protocols/ action plans for particular sites, for same type of sites, customers have different preferrences: eg. Car dealerships, construction sites, metal fabricators,...
They also need to properly address how to respond when presented with a situation (person present) is it a worker, cleaner, customer, gate left unlocked, suspect,... or camera equipment issues: (high false positives, connectivity issues, ...
Do they assess properly the action to do : call the police, call a guard, go silent, activate talk down, audibles/ strobes, call the customer, call IT, call manager, ...
Creating scenarios to test behavior of the operator and learning is not simple, good testing can eliminate operator errors that can have huge impacts. Don't know if the software helps but there is a problem that needs solution.
I am not questioning the need for training... nor that site-specific differences mandate specialized content to be mastered by new surveillance operators. As a tool to learn procedural-based responses, I've got no issues with it. Simulators are pretty good at teaching basic SOPs using repetition and simple memorization techniques. Simulators can help accelerate the learning curve in relation to response... which is certainly important - but only half of the equation.
My problem with AI-based simulators for surveillance operators is that I don't think this same acceleration principle is true for learning behavior - the equally important other half of the job. Some might say the more important half.
AI-based cartoon robots are inadequate at imitating a real human with possible criminal intent. Right-clicking a crowd of robots and choosing 'protest' is not going to produce the stress levels, chaos and confusion of a real-life crowd incident.
Simulators work great - if they are actually 'simulating' - as close as possible - something real (like flight simulators and other virtual reality devices do).
In my opinion, identifying and reacting to the subtleties of human behavior is much harder to learn via simulators, primarily because of the difficulties in emulating anything other than the most obvious scenarios (that don't really need training to identify; i.e. man waving a gun around on the street).
I believe that it would help for city monitoring, but for other operations such as casinos, stores, malls etc. It would be a different story. Unless they have different types of Sumulators for different types of operations.
From my understanding from the website, they can build a program/scenario to fit the needs of the end user.
Just watched the second video without sound... hilarious. set to the right music it could be the trailer for Will Smith's next end of the world summer blockbuster.
I think the concept is pretty smart; but as an outsider to the "operator" world, it seems relatively useless. The first thing I didnt seem to get (maybe because I was watching without sound) is it doesnt seem like youre operating the actual VMS that you have deployed, which would seem to be a bit of a weakness when youre attempting to train someone. A better system, it would seem to me would be to have some sample video generators running into a video encoder thats connected to your actual VMS (patent pending, not really). The second thing is I just dont think there is enough graphic detail available to train on anything other than basic traffic patterns/flow. I would think you would be better off just training with canned video of events from your actual building so that you also learn any camera blind spots, glare issues at specific times of day, and any other specific to your location problems. The only thing you would lose training wise is any optical PTZ control training, which you should just buy and old Pacman arcade game and stick the person in front of that for a week for those motor skills...
I finally got a hold of someone at the company. You can find more detail about the simulator here.
Newest Discussions
| Discussion | Posts | Latest |
|---|---|---|
|
Started by
Lynn Harold
|
5
|
1 minute by Lynn Harold |
|
Started by
Mark Rindy
|
24
|
less than a minute by John Honovich |
|
Started by
Undisclosed End User #1
|
6
|
less than a minute by Jeffrey Stout |
|
Started by
Undisclosed Integrator #1
|
13
|
less than a minute by Dennis Ruban |
|
Avigilon Or Inaxsys?
(29)
Started by
Undisclosed End User #1
|
29
|
about 2 hours by Undisclosed End User #1 |