Subscriber Discussion

Statistics Released For Video Verification Monitoring Center

I believe strongly that video surveillance linked to security systems is the key to eliminating false alarms to police. A lot of waste is going on with traditional alarms - some police departments actively measure it and the results range from 95~99%. If you can eliminate false alarms to police, they will respond faster to your clients sites. If they respond fast and if they are silent alarms you will get apprehensions. There is not much published data on this subject.

I just completed analysis of events at our video verified outdoor sites for the months of Sept and Oct and wonder what readers opinion of my data might be. 1/4 of the sites were Car Dealerships, 1/4 were construction sites, 1/2 were out door storage areas. The car dealerships were open lots, otherwise were fenced properties. 3/4 of sites were triggered by video analytics, 1/4 were videofied cameras.

A. No one seen, wildlife, weather, end of daily tour, maintenance issues, power outage, testing: 4,615 events 46.14%

B. Person Seen - No threat perceived: 5,369 events 53.68%

C. Suspcious person - police dispatched - arrests: 2 events 0.02%

D. Suspicious person - police dispatched - no arrests: 0 events

E. Suspicious person - police dispatched - false dispatch: 0 events

F. Suspicious person - Alarms activated by operator - left premises - None/Guard/or Ref sent: 16 events 0.16%

The challenge I have seen with audio verification schemes in the past is that the monitoring central station rarely wants to risk the liability of ignoring a legitimate alarm, even if the audio tells them that no one is on the premises. To cover themselves, they always dispatch the police anyway. The thinking is, is better to have 99 false alarms than it is to miss one real burglary. (Particularly if the owner is paying any fines.) This CYA attitude largely negates the value of audio verification.

I wondering if this same attitude might prevail with video verification? Unless 100% of the premises is covered with a camera, will central station personnel want to assume the liability of ignoring an alarm when they see nothing on the video?

I can see video verification working in the outdoor applications that you describe, but how about a building with multiple rooms and many places for a burglar to hide?

Robert, thanks for sharing. I am trying to understand your data, specifically category A: "No one seen, wildlife, weather, end of daily tour, maintenance issues, power outage, testing" and B: "Person Seen - No threat perceived"

Are you saying that without video verification, those 10,000 events would have triggered police dispatch? Or are these systems increasing the false alerts (e.g., category A)?

My other question - there were 18 suspicious persons but evidently 10,000 events had to be watched to find those 18. What's the cost and value of looking at 10,000 events to track down 18 potential security risks?

The main areas to be covered are the doors. The goal is to eliminate false alarm dispatches to police. If you do video verification it needs to comply with the standard CSAA-V02-2012, section 3.9 requires <100ms between activation of the zone that causes the alarm and the video clip sent to the central station. Entry door activation needs to be considered and the video clip needs to be sent only if the panel goes into alarm. We have a 10 sec clip and it starts 5 sec before the door is opened. Motion and glass break would be instant. Hiding doesn't seem to be an issue for us.

Category A includes the following: A.1 No bounding box, normal tour closing = 2578, A.2 Weather/ wildlife/ reflection = 1532, A.3 ~ A.8 Maintenance, testing and troubles = 505

Category B includes the following: B.1 Person seen - no threat perceived = 3916, B.2/3 Auth person/vehicle on site = 620, B.4/5 False trip by person/vehicle adj to ROI = 829, B.6 Bypass Request by user = 4

These systems create more events than motion detectors but if the operator is uncertain what triggered then they will make more dispatch calls than video analytics as we can see what exactly tripped the camera (bounding boxes). We are happy with the result. It is a guard replacement service - the customer obviously has a high loss rate without protection.

Our Mom & Pop retail establishments are secured by perimeter alarms and video monitoring. Although video monitoring at all locations has probably paid for itself on two separate occasions over 5 years, there have been several instances (we don't keep records but it's been more than 2, less than 10 by memory) when video has saved us police calls initiated from ADT false alarms.

After ADT perimeter security installation at one site, until recognized and repaired, one door created false alarms because it had too much "bounce" from rapid pressurization changes. On several very gusty nights, we got alarm calls from ADT. Since camera coverage is complete, we were able to confidently verify no activity.

While avoided false alarm charges haven't paid for that monitoring system by themselves, when you add in the cost of us going to the sites in the middle of the night (1-2 hour round trip), I think we can chalk up another win for video.

An interesting aside: ADT installed the video and charged us $70/month for video monitoring. ADT installed the perimeter security and charged us $60/month for perimeter monitoring. When we received these alarm calls, we asked ADT what the video showed, but they indicated they couldn't tell us because they weren't monitoring it. Imagine it took a false alarm for us to learn that, so we promptly cancelled video monitoring. This was probably more than 4 years ago, so we probably have saved more than $3,000 in video monitoring fees since then (still not completely paying for the install, but it's all goodness). I would encourage anyone who has some form of integrated video monitoring to create a few test cases to verify you're getting what you're paying for.

Random question, did they ever fix the door contact? Kind of easy to compensate for the bouncing...

Yes, the handyman took care of it. It wasn't a switch debounce. The locked door had too much free play and so the magnet moved in and out of the reed switch's closure range. From a maintenance perspective, we should have had it fixed anyway.


Why can't the video monitoring system be filtered to avoid the Normal Tour Closing alerts in A.1 and all of the alerts from A.3 ~ A.8?

It doesn't do anything to avoid false dispatches to police, but it could drop a portion of the total alerts that have to be verified.

Marty, perhaps there is a misunderstanding of the normal tour closing alert event. The monitoring software will generate an alarm event based on a schedule set up for that site to trigger the operator to perform a tour of the cameras, usually once a day because we rely on the analytics. The operator then opens the site and performs the tour programmed for that site. When they complete the tour they indicate the tour outcome, which should be "no one seen" because the analytics should have detected them. I included the number just for completeness.

Robert, Your operators than watched 10,000 video clips / looked in 10,000 times over that period? How does the economics of that work? Does a customer pay per event / look-in? What type of costs does that run? I am trying to understand the cost of this service vs the loss rate without protection?

John, It is a big number for sure, there were some issues in October, however, I would expect about 3000~4000 events/per month. The customer pays a flat monthly rate. I watch monthly the effective hourly rate for operators handling all the events. I would rather not say how much exactly, but this is a guard replacement service. They are use to paying $3,000~$8,000 per month for a guard at each site. Depending on the type of event: weather duration may take 2 seconds, for an apprehension we are assisting police on site over the phone for maybe 20 min. Indications are that the number of suspicious persons alerts is not growing linearly as more sites are coming on line. In fact it appears to be declining. I believe the apprehension rate is creating a deterrent effect.

FYI - I notice that when I post a reply to a person as opposed to the main heading; the web interface hangs up with a "loading..." Also I still am unable to post video or images to ipvm??

Well, they had to installed adequate door contacts then -- with wide gap.