New to ZM. This weekend installed ZM on Centos 7 as a VM. After some fumbles, got ZM going with a HIKvision camera. Camera pointed to a hall way, indoors, for testing.
That seemed to work well enough to begin to get familiar with ZM (and CentOS).
Decided to point the camera outdoors, through a window for additional testing. Window is closed, as it is cold still, so the camera looks through glass *and* a dark fiberglass screen. Monitor looks fine so I thought to test motion detection.
Seems to not trigger "most of the time" Once, it did trigger and record, but the playback did not show anyone in frame. I know there was, as I was the test subject and had a car in drive as reference.
I wonder about looking through, but, since the image looks pretty good, doubt it has much effect on motion detection.
Hints?
New install, motion detection not reliable
Re: New install, motion detection not reliable
Added a Zone, mostly by luck, and seem now to be triggering. I'll continue to pull on my boots, but any hints are welcome.
-
SteveGilvarry
- Posts: 494
- Joined: Sun Jun 29, 2014 1:12 pm
- Location: Melbourne, AU
Re: New install, motion detection not reliable
Written for someone else of course, but might help. 
https://wiki.zoneminder.com/Understandi ... or_Dummies
https://wiki.zoneminder.com/Understandi ... or_Dummies
Production Zoneminder 1.37.x (Living dangerously)
Random Selection of Cameras (Dahua and Hikvision)
Random Selection of Cameras (Dahua and Hikvision)
Re: New install, motion detection not reliable
I'd found that many of the ZoneMinder presets for motion were too 'squelched' to be effective in my case. My first ZM system was one I had to install at a remote site in a day and then leave. All cameras were looking outward from the main building. A few days later I was wondering why I hardly getting any events. Changed some of the parameters and then things started working.
Specifically, I've have most of my cameras set to 'Min' alarmed/filtered/blob areas of less than 1%. (You have to switch to pixel mode, type in a number and then go back to percent. ZM doesn't seem to want to allow you to enter values of 0.61, for example.) A lot of factors go into whether or not you'll get reliable motion detection. Some of my areas of interest cover from 0 to 15 meters from the camera. With a wider angle lens, the person standing 15 meters away is awfully tiny - pixel wise - so that's why I'm down below 1% in cases.
You're also having to factor in pixel threshold and even things like the colour of clothing the person is wearing. Someone wearing a white parka in a snowstorm won't look like much to ZM, whereas a guy in a red parka would. What may stand out to visually to you and me may not quite make the cut with ZM, as its filtering process can reduce the 'area' of the person to less than your minimum, and you don't get any recording.
Day/night is also a potential factor. My cameras switch to IR mode and are in black and white at night. The neighbour's dog can wander by during the day and it stands out enough from the background to trigger motion. At night, where everything is just shades of grey, it doesn't trigger anything unless it's really close to the camera.
The only thing I can think that being behind a window would influence would be the effectiveness of any IR LEDs on the camera at night. Daytime - as long as there aren't strong reflections - should be okay.
One of the things I like about ZM is the ability to infinitely configure the system to my needs. This also means of course a lot of tinkering with settings to get things right. Start with one of the built-in presets and work from there. If you're not capturing motion, expand the min/max areas until you do. When you get a few events recorded, go through and look at the info that ZM captured regarding alarm and filtered pixels and make further refinements.
Here's the settings from one of my cameras as an example. This is a 3MP 2048x1536 camera, although I'm using the second stream to analyze motion at 640x480, with a lower frame rate. Units are in percent.
Alarm Check Method: Blobs
Min/Max Pixel Threshold: 40
Filter Width/Height (pixels): 5 / 5
Zone Area: greyed out, but showing 54
Min/Max Alarmed Area: 0.68 / 45
Min/Max Filtered Area: 0.61 / 40
Min/Max Blob Area: 0.61 / 0
Min/Max Blobs: 1 / 0
Overload Frame Ignore Count: 2
Extend Alarm Frame Count: greyed out
...so there must be at least one blob, 5x5 pixels in size, and must take up at least 0.61% of the space, and the pixels must have changed by 40 (of 255 levels), and the 'motion' can't be bigger than 40%.
Hope this helps.
Bill
Specifically, I've have most of my cameras set to 'Min' alarmed/filtered/blob areas of less than 1%. (You have to switch to pixel mode, type in a number and then go back to percent. ZM doesn't seem to want to allow you to enter values of 0.61, for example.) A lot of factors go into whether or not you'll get reliable motion detection. Some of my areas of interest cover from 0 to 15 meters from the camera. With a wider angle lens, the person standing 15 meters away is awfully tiny - pixel wise - so that's why I'm down below 1% in cases.
You're also having to factor in pixel threshold and even things like the colour of clothing the person is wearing. Someone wearing a white parka in a snowstorm won't look like much to ZM, whereas a guy in a red parka would. What may stand out to visually to you and me may not quite make the cut with ZM, as its filtering process can reduce the 'area' of the person to less than your minimum, and you don't get any recording.
Day/night is also a potential factor. My cameras switch to IR mode and are in black and white at night. The neighbour's dog can wander by during the day and it stands out enough from the background to trigger motion. At night, where everything is just shades of grey, it doesn't trigger anything unless it's really close to the camera.
The only thing I can think that being behind a window would influence would be the effectiveness of any IR LEDs on the camera at night. Daytime - as long as there aren't strong reflections - should be okay.
One of the things I like about ZM is the ability to infinitely configure the system to my needs. This also means of course a lot of tinkering with settings to get things right. Start with one of the built-in presets and work from there. If you're not capturing motion, expand the min/max areas until you do. When you get a few events recorded, go through and look at the info that ZM captured regarding alarm and filtered pixels and make further refinements.
Here's the settings from one of my cameras as an example. This is a 3MP 2048x1536 camera, although I'm using the second stream to analyze motion at 640x480, with a lower frame rate. Units are in percent.
Alarm Check Method: Blobs
Min/Max Pixel Threshold: 40
Filter Width/Height (pixels): 5 / 5
Zone Area: greyed out, but showing 54
Min/Max Alarmed Area: 0.68 / 45
Min/Max Filtered Area: 0.61 / 40
Min/Max Blob Area: 0.61 / 0
Min/Max Blobs: 1 / 0
Overload Frame Ignore Count: 2
Extend Alarm Frame Count: greyed out
...so there must be at least one blob, 5x5 pixels in size, and must take up at least 0.61% of the space, and the pixels must have changed by 40 (of 255 levels), and the 'motion' can't be bigger than 40%.
Hope this helps.
Bill
Re: New install, motion detection not reliable
Thanks gents. The settings were . . . unsettling me. Still are, but less so thanks to your inputs. I am getting better results now.
The through the window bit becomes problematic come night fall as the video becomes unusable due to reflection of the IR LEDS. When I get a bit more comfortable, I will mount the cam outdoors and perhaps get another. Uh, oh, he's getting brave now . . .
The through the window bit becomes problematic come night fall as the video becomes unusable due to reflection of the IR LEDS. When I get a bit more comfortable, I will mount the cam outdoors and perhaps get another. Uh, oh, he's getting brave now . . .