Videoiq False Alarms And Missing Suspects

What you think about Videio IQ Video Analytics ??

I Have concern about missing suspects & false alarms.

Where are you using VideoIQ analytics? What's the scene? What's the camera? What subjects are you missing?

Overall, and relative to other video analytics, VideoIQ does well. See: 2014 Video Analytics favorites and Testing VideoIQ Rialto

That's not to say it won't have issues. If you can share where you are having problems, we can help provide feedback.

In a private residence, the site is huge, the aim is to monitor the perimeter are of the house. The site has lots of trees, cats and birds, cats might access the site easily, the customer has also dogs. The camera is the hd model videioiq. When we lower the sens the system is missing people . Of we increase the sens we get false alarms, sometimes the dogs are activating the system. We have been trying to solve the issue for the past two months. We ate just tired , i think we had to offer a permitter pir, beam or otgher technology but not analytics.

Karim, when you lower the sensitivity just enough to reduce false positives(alerts without sufficient cause) to an acceptable level, are the false negatives (sufficient cause without alert) all of a similar type, e.g., small person in driveway, or does the system appear to 'break' everywhere at the same time? Is it related to particular time of day or camera? Have you optimized the camera settings themselves? How much is cost a concern? Is external lighting a possibility? As for being tired right after eating, this is not unusual, I recommend a short but brisk walk.;)

Here is a link to another discussion exploring a wide-range (maybe too wide) of perimeter protection/intrusion detection options. And here's a document you may have seen already, its VIQ's perimeter protection for small sites pdf.

And are PIR and beams going to be immune from the animals on site?

The issue seems to be that you have a very complex scene which will be challenging for most analytic systems. I am not sure there is an easy solution.

This sounds like the eternal complaint about video Analytics. Dissecting the complaint without seeing the sight would make me question....

How far from the camera are objects supposed to be detected?

How close can objects like cats get to the camera?

Is there EVEN and ADEQUATE lighting?

IF you used other types of detection would you be covering the same area, expect better results?

I've seen good results from the VideoIQ product but managing expectations and proper design saves time and heartache. Just my 2 cents worth having been in the video analytic industry.

Managing the relationship between PD / FAR is always a challenge. Where VIQ excels is the time spent doing initial tuning and seasonal changes.

This is the site, as you can see it is quite huge and located in an open area. Videio IQ are remotly checking the configuration settings. but if we had to connect remotly the systems with a control room it had been a different story, i beleive that we can improve the system by making sure we have sufficiant lighting during night time and also close more the field of view.

what is the max field of view you recomend ?


VideoIQ has FOV recommendations in their design docs. I would personally say you probably don't want anything over 200' wide in this case.

If you've adjusted sensitivity and can't find a sweet spot that reliably detects humans while not detecting dogs, you likely do need more lighting, or more pixels, or both.

From a Google perspective I can see difficulty at several potential issues. As a note, in my past I have installed a $465,000.00 residential alarm system. Did cameras get installed on the residence looking out? Are they installed on the fence line? When you say you will add lighting are any facing east/west where sunrise and sunset will be an issue? What is the annual climate? What is the intended use? Is there on-site security (command post) or is it all remote to a central dispatch? Images from the prospective camera locations would be useful. But I wouldn't post them here. Last but not least and a complete product plug, shamelessly stated.....this is what VIQ and FLIR do together when identification isn't important and detection is. You need the camera far enough from cats and dogs so they don't become Godzilla to the camera at times. Not so far that humans look like cats and dogs. Bushes and trees create different shadows and shadows have to be managed. Shadows change based on sun travel and natural growth so they are seasonal. You will not get a 100% detection and 0 false or nuisance alarms on this site with anything short of ground clearing a path around the property and putting up dual fences with a clear zone. AKA "prison perimeter". You can increase your odds with help though. John H asked if you thought perimeter detection like PIR/MW detectors would also discern or discriminate cats, dogs, crawlers and trees to prevent you from having too huh of expectations. Sites like these have been done with success after gaining experience, managing all expectations and proper design with everyone at the table. Just for humor I will tell you one experience doing these. We installed a couple of thousand feet of dual cable leaky coax with everyone's knowledge. Over a 3 day break the owners wife had 40' palm trees installed in 6 locations, directly cutting out cables as though they were the guide lines. Thousands of dollars and delays because we didn't work with "her" landscape architect on the security design and did a "need to know" with the contractor. Anything can happen.