Subscriber Discussion

Why Can't You Use A Saved Clip To Set Motion Detection Sensitivity?

U
Undisclosed #1
Mar 22, 2017
IPVMU Certified

Like when something in the middle of the night (noise, wind, headlights etc) is occasionally causing irrelavent motion to trigger an event.

You need to adjust sensitivity or other parameter, but can only guess because it's not the middle of the night, and all you're able to use is the live view to preview with.*

You should be able to just put the clip from the event in a loop and fine tune using that.

What about doing the ridiculous and putting a 7-in spot monitior (with the clip looping on it) right in front of the camera you are trying set?  ;)

*Actually I don't know if any VMSes support this today, but I don't think it's in ONSSI and some  others.  Maybe I'm wrong, though.

U
Undisclosed #2
Mar 22, 2017

Motion detection sensitivity is generally set in the administrative application of a VMS, while recorded video is displayed in the client application.

I don't know of any VMS's that display recorded video in the administrative application - which would be a requirement to do what you are proposing.

Also, why would you use a middle-of-the-night motion spike to set a cameras overall motion sensitivity?  Wouldn't this just make the day-time sensitivity too low to be effective?

U
Undisclosed #3
Mar 22, 2017

A number of VMSes support playing back video files as loop. So you can do exactly this. But even if the VMS doesn't support this, you can set up VLC to stream video via RTSP, and it can be set to be looped. So you can add that RTSP stream as a camera and do exactly what you're looking for. 

(1)
U
Undisclosed #1
Mar 22, 2017
IPVMU Certified

Thanks.  Yes, you could restream a clip back.  

My question is why don't they just let you do it directly?

How can you ever know with any certainty that your adjustment to VMD to eliminate an occasional but repeated false alarm will be successful, before it happens again.

U
Undisclosed #3
Mar 22, 2017

Because they haven't gotten a level of demand to justify the effort. For VMS manufacturers, developer time tends to be a precious resource. So it's allocated towards things that will either generate revenue such as a project that requires a feature, something that will generate marketing interest, or the pet feature of whoever is in charge of the developers. Once all of those are done, there is time for other features. 

For a lot of VMSes this feature wouldn't be hard, but it would be a bit time consuming. A lot of the parts already exist, but getting them to play together might take some effort.

U
Undisclosed #1
Mar 22, 2017
IPVMU Certified

So such a feature would have never been helpful to you?

U
Undisclosed #3
Mar 22, 2017

There have been times when it would have been useful for sure. But the number of times I've needed it hasn't been super high. And given I can do looped video on just about any system, it wouldn't be the thing that makes me choose one VMS over another. So I can't build a strong business case. 

That doesn't make it a bad feature idea. It's just that no one has presented to a VMS company a compelling argument that making that feature will make them money. An idea can be good or useful without being an idea that is worth investing money in. 

If you feel strongly about it, reach out to your VMS manufacturers of choice and ask for it. Your best chance of getting it quickly is to be able to present a solid business case for it. Something like If this feature is implemented, then we can be the only VMS that fulfills the requirements for Project X which is worth $Y. Or we have made X service calls over the last Y months costing us $Z and having this feature would have eliminated it. 

Or you can just ask for it. VMS manufacturers will listen, and add it to the list of suggested features that might get added as busy work for developers. But keep in mind that the number of suggestions they get is pretty high. And a lot of them are good ideas. So it might happen, just don't hold your breath. 

 

Avatar
Josh Hendricks
Mar 22, 2017
Milestone Systems

This is a great idea and it wouldn't only be useful for tuning midnight motion detection. Some cameras see very little motion even during business hours and it can be time consuming waiting for something to happen to verify the settings are right.

Most of our customers are using the automatic motion detection settings, but frequently things need to be customized so this would come in handy. While I'm sure it's been mentioned internally before, but I'll give it a bump anyway.

U
Undisclosed #2
Mar 22, 2017

I'm still not sold on the case use for this.  I don't get why this would help you set up your VMD settings appropriately so they would work both in even lighting (daytime) and sporadic lighting (night time) situations.

 

I don't understand how a looped playback clip from night time can help you configure a 'universal' (i.e. all the time) VMD setting (rather than using 2 different schedules to use higher sensitivity for VMD during the day, and a lower sensitivity setting for night time).

 

Avatar
Josh Hendricks
Mar 22, 2017
Milestone Systems

I'm with you on the benefit of having two different motion detection settings based on a schedule. But even for products that might have that or if they were to add this, there is a benefit in timeshifting so that you can avoid having to be on site or login remotely at ungodly hours just to create an after-hours motion detection profile.

If I were to prioritize this myself, I'd probably put it somewhere toward the bottom of the list as a nice-to-have. but I still see value in it.

For Milestone customers, it seems like the automatic motion detection is "good enough" at filtering out image noise so the noise at night might not trigger as much false motion as it used to.

(1)
U
Undisclosed #1
Mar 23, 2017
IPVMU Certified

I'm still not sold on the case use for this.

To be sure, this isn't really about multiple motion settings for day/night; it's addressing the more general case of adjusting the parameters to their ideal settings when trying to account for occasional events that are not easily duplicated.

Maybe you could explain how you personally would deal with this scenario (for simplicity assume I only care about daytime):

  • There is a large neighborhood cat that every so often (once  a week?), triggers the motion detection that I have on the walkway to my house.
  • I would like to ignore the cat without affecting the triggering of larger objects. 
  • So everytime it happened I bumped the sensitivity down a little bit until it didn't trigger anymore.  I then had to test with larger objects to make sure it was still sensitive enough to get those.
  • Needless to say this took several weeks/events to get right.

If I could have just set the sensitivity based on playing 2 clips, one of the cat and one of a person, I could have tuned it better in less than 5 min.

So how would you have done it (without changing equipment)?

Or more generally, how do you ever know how to optimally set the sensitivity of motion detection when the event that triggers them is not happening at the moment you are setting them?

Rent a cat?

(1)
(1)
U
Undisclosed #2
Mar 23, 2017

My point is that a 'properly' set VMD daytime setting will always be plagued with the false alarms of the night time variety - based largely on the general unevenness of lighting that occurs.  i.e. headlight beams that fall in a field of view will change pixels, even if the headlights or the vehicle itself is off screen.  Motion event.

"it's addressing the more general case of adjusting the parameters to their ideal settings when trying to account for occasional events that are not easily duplicated."

VMD is the weakest of analytics - it simply notes pixel change.

Regardless of how or why this pixel change occurs (shadows, cats, etc) VMD is unable to distinguish anything beyond the fact that pixels have changed.

 

  • "So everytime it happened I bumped the sensitivity down a little bit until it didn't trigger anymore. I then had to test with larger objects to make sure it was still sensitive enough to get those.
  • Needless to say this took several weeks/events to get right."

I do not accept this conclusion.  There is no 'right' that can cover both.  You can only 'bump down' the sensitivity until it ignores anything the same size or smaller than the cat.  As VMD sensitivity is basically lowering or raising the 'percentage of pixel change' levels, this solution fails when a 2% larger cat creeps down the walkway to your house. 

To avoid the neighborhood cat - or any of his heftier buddies - from setting off a motion alert, you would need to teach the system what a cat is first, and then command it to ignore cats.

The obvious solution to the creeping cat dilemma is video analytics.  Not sure if there are cat-specific VA though.

I don't think there is an answer to the OPs dilemma unless he/she could somehow teach the system to ignore whatever it is that his/her system is alarming on during those off hours.

In my experience, VMD sensitivity is 'tuned' as best as possible taking into consideration resolution and how wide the FoV is - and lighting.  If lighting changes throughout the day and night then you need different VMD sensitivity settings (i.e. schedules) if you don't want the unevenness of night time lighting changes to set off motion alerts based on your 'correct' day time settings.

U
Undisclosed #2
Mar 23, 2017

sorry, I missed your caveat that you don't care about night for the cat dilemma.  but this is what the OP was mentioning.

U
Undisclosed #1
Mar 23, 2017
IPVMU Certified

You can only 'bump down' the sensitivity until it ignores anything the same size or smaller than the cat.

That's it right there.  

That's what I needed to do, make the VMD not trigger on a small animal, but still trigger on a small human.

And I did it exactly as you say by 'bumping it down (or up)' until it is what I consider optimal.

It just took weeks instead of minutes.

The optimization of motion detection sensitivity to set the triggering threshold of an event, i.e. get this/skip that, is common, no?

Again how do you set your motion detection sensitivity/threshold properly, using the VMS controls, when the expected event(s) are not currently in the live view?

Please answer this.

U
Undisclosed #2
Mar 23, 2017

"how do you set your motion detection sensitivity/threshold properly, using the VMS controls, when the expected event(s) are not currently in the live view?"

The same thing that integrators have been doing for decades when configuring motion detection sensitivity - you do what you can to imitate situations that will occur, but that are not happening when you are there.

i.e. one person sits at the controls while another person (generally referred to as 'the flunky') wanders around at various depths in the FoV while the person configuring settings views and adjusts sensitivity based on readings generated by the flunky.

It is a science and art combination.  The more familiar you are with one platform (i.e. your go-to VMS), the more precise you can get based on experience.  However, there is never a perfect (static) solution in an outdoor FoV when using VMD to trigger alerts.  You will always have some false positives when the environment in the FoV changes (weather, lighting, cats, etc) over time.

U
Undisclosed #1
Mar 23, 2017
IPVMU Certified

i.e. one person sits at the controls while another person (generally referred to as 'the flunky') wanders around at various depths in the FoV while the person configuring settings views and adjusts sensitivity based on readings generated by the flunky.

So I agree that what has been done for decades.

I'm not sure why you are not then seeing a valid use case for clips when events are not easily imitated or anticipated by a flunky; animals/middle of the night/glare issues etc...

What do you pay your flunkies ;)

U
Undisclosed #2
Mar 23, 2017

"What do you pay your flunkies"

flunky scale.

I guess I can see your point regarding the use of recorded clips instead of flunkies to 'tune' settings...

clearly, removing the need for a flunky is a good idea - and as Joshua noted, RE: Milestone's 'automatic VMD', that is their solution to remove the need for a flunky.

I was focused on the fact that one universal setting - for all times of day and night - will never provide a 'just right' result. 

 

U
Undisclosed #1
Mar 23, 2017
IPVMU Certified

flunky scale.

Do you actually use real flunkies?  

These days most people have been trying to save a buck or two by using lower labor grades, like grunts, lackies and newbies ;)

 

(1)
Avatar
Ari Erenthal
Mar 23, 2017

Nothing like a genuine, Acme brand, Hired Goon. 

UI
Undisclosed Integrator #4
Mar 23, 2017

I can see value in this by tuning to clips that created alarms day or night. 

I also see where it would only work with server based motion detection. 

U
Undisclosed #5
Mar 23, 2017

Inevitably variables change. If you are looking for absolute numbers that are always the same and unchanging then use a video clip.

I would rather tune the cameras to each unique view they are responsible for recording.

Smart analytics that narrate the efficiency of recording are coming. Soon.

 

New discussion

Ask questions and get answers to your physical security questions from IPVM team members and fellow subscribers.

Newest discussions