Sensitivity/reliability of motion detection at a distance

[DKtucson 0]

OK guys.. here's the situation:

Residential install--backyard scene. Dome cam mounted under eaves of porch with 3.6mm lens getting a end to end shot of a kidney shaped pool.

Internal IR in the dome is disconnected and we are using a IR floodlight that pumps pretty good over the 60ft distance--photo attached.

The motion detection at the far side of the pool isn't 100% reliable.

But the again what motion detection is..? anyways here is what we've been running into and what we've experimanted with:

1. It seems that if we set thershold too low for detection it just goes off from minute fluctuations of light

2. Bump it up one more increment from 6 (false alarm zone) to 7 gives about 75-80% reliable motion pickup depending on area

3. Put a flashlight in the hand of the test prowler--LED or mini-mag and bingo they register every time. This is Tucson Az we are talking about..NOBODY is going to be walking out in the darkness of night in the desert without a flashlight

4.If they come into the foreground where they take up a larger area --bingo--they register

We tested an additional emitter off to the left of the scene shining accross and that also helped in the area of the slide/far wall.

 

Here is my philosophy and any input is appreciated---One should not depend on software to adequately do the job of correct hardware.

The client wanted a wide view of the pool area--OK

The client didin't want to spend a fortune on cameras--the budget was sub-$2000 for 8 cams , cabling and a dvr. Needless to say that does not pay for WDR or starlight cams--

 

So, the wide angle view of the pool makes for a very small target in the far distance--we simply have to supply more light to the area so that a dinky distant person registers.

 

The client is expecting perfection 100% of the time--but as with any sensing system there are trade offs for sensitivity/reliability.

To draw an anology my car might get pelted with a large bird dropping--no alarm. So This bothers me as someone *could* egg my car and I wouldn't be alerted. So I adjust the sensitivity up and toss an egg at the car---woop woop woop. Kewl, I can sleep better at night....um except for the fact that a passing Harley, VW Bug or Bass Kicker is setting off my alarm and waking me up

[Soundy 1]

You can't fault "software motion detection" entirely - different DVRs implement it differently with different degrees of accuracy and different levels of "adjustability".

 

Some now include video analytics, which does require more setup, but is far more precise: for example, it lets you pre-define the sizes of people, animals, vehicles, etc., and will let you define directions of motion, so for example, an alarm could trip when a person enters the yard but not when they leave it, or trips when a person hangs around in front of the gate for more than 30 seconds... or whatever.

[DKtucson 0]

Hi soundy,

Thanks for the reply. I did some experimenting and the old "is it hardware or is it software?" scenario. I installed another vendor software WebcamXP and set it up--it detects motion flawlessly or a helluva lot better than the original software.

The drawback is that while it has good features in column A--it's missing one in Column B.

The original software had the feature where you could do the following:

1. Install a DVR at your business

2. Install a DVR at your home

3. Install at a 3rd location and bring up the other 2 systems as net connect and have redundant recording seperate from the original DVR's

 

The problem with FTP upload with everyone's software is that all they upload is a sequence of stills

 

Here is a still from the Pool scene at night--bird caught flying across--circled in red

[bpzle 0]

Maybe I missed this somewhere... but what's the problem with setting it on the "sensative" side? What's the big deal if it records something of no importance? At least your sure to get what you do need...

I'm 100 percent for customer satisfaction, but we can't change the capabilities of the equipment. If he wants a more reliable form of motion detection, he needs to be willing to pay for equipment designed to do what he wants. Plain and simple.

 

The only way I could see it being a PITA is if you had alarm events triggered by video motion detection. But we highly discourage that anyway... especially outdoors.

[DKtucson 0]

Hi BP--thanks for the reply

Here is the customer's situation:

She takes lengthy vacations and came back home last year to a $3000 water bill. No signs of leaks/running toilets etc. She's not going to be home per se to view on a regular basis. We want it to record only on event detection as she has 8 cams going to a TB drive.--lots of storage but not a bottomless pit.

 

In order for it to be sensitive enough at night we're risking more false records during the day--that's for sure.

 

And I'm with you on "you get what you paid for" philosophy.

[bpzle 0]

Ahhh... I see.

 

What kind of DVR is it?

 

Have you thought of reducing the FPS/ compression quality to maximize storage?

[Soundy 1]

In this sort of situation, masking may be more important that simple sensitivity - for example, if there's a busy highway in the background of the frame, you want to be able to block out that area so it's not triggering recording every time a car goes by.

 

Keep in mind that sometimes you don't have to have specific areas unmasked, either - for example, if masking is in "blocks" and the top of a fence cuts through the middle of a row of blocks with the busy highway beyond... you can safely mask out the entire top of the fence, as anything coming over it will still trigger the unmasked area below. As another example, I have instances where there are tables on a restaurant patio and then a road in the background... I find I don't need the tables themselves fully unmasked if the road is also caught in those areas, because the lower part of the table still catches people's legs moving, or the people shuffling in their chairs, and that can be used to trigger recording.

 

When setting masking, in other words, look at the whole scene and think intelligently about what areas actually need to trigger and which you can mask out.