Motion detection problem - a bug I think

[maciekc]

Please take a look at those pictures
http://dig.hopto.org/zm/001-capture.jpg - that's frame one frame before event
http://dig.hopto.org/zm/002-analyse.jpg -that's move I made
http://dig.hopto.org/zm/007-analyse.jpg - that's my position after move

I'm not moving anymore but it keeps recording:
http://dig.hopto.org/zm/008-analyse.jpg
http://dig.hopto.org/zm/009-analyse.jpg

After I make the same postion as in first frame it stopped recording
http://dig.hopto.org/zm/031-capture.jpg

It always checks wih 001 - 008 with 001, 009 with 001.

In my oppinion it should check motion beetwen frame 008 with 009
next 009 with 010

Can I change that somewhere?
I hope you understand what I wrote if not please tell I will try to explain it better.
It is very important for me.

Here are my settings:
Alarm Check Method=Blobs
Minimum Pixel Threshold (0>=?<=255)=30
Maximum Pixel Threshold (0>=?<=255)=255
Minimum Alarmed Area=600
Maximum Alarmed Area=76032
Filter Width (pixels)=7
Filter Height (pixels)=7
Minimum Filtered Area=600
Maximum Filtered Area=76032
Minimum Blob Area=600
Maximum Blob Area=65535
Minimum Blobs=2
Maximum Blobs=65535

[maciekc]

I got an idea to make the detection better.
I explain it on example:

For value 1 it should check:
1 with 2
2 with 3

For 2:
1 with 3
2 with 4
3 with 5

For 3:
1 with 4
2 with 5
3 with 6

For 4:
1 with 5
2 with 6
3 with 7

Understand what I want to get?
If not tell me and I`ll post images which makes it recording without ending it.

[zoneminder]

The motion detection algorithm checks each frame against a reference image. This reference image is a composite of previous frames with each frame blended in at a certain percentage (Reference Image Blend %ge in the Monitor setup). So the reference image at image 'n' is 10% of n-1, 9% of n-2, 8% of n-3 etc. This is to allow the reference image to adapt to gradually changing conditions, e.g. daylight changes to nighttime etc. This percentage is configurable for each monitor, so if you want a more static reference image then use a lower percentage, fi you want a more rapdily adapting image increase it.

You can also modify the behaviour by changing ZM_BLEND_ALARMED_IMAGES in Options->Config. The excerpt from the help explains why,

To detect alarms ZoneMinder compares an image with a reference image which is formed from a composite of the previous images. This option determines whether images that cause events are included in this process. Doing so may increase the precision of the alarmed region but can cause problems if wholescale lighting changes cause alarms as this would not get fed back into the image and an alarm may persist indefinately. A better way to achive the same effect in most cases is to lower substantially the reference blend persentage in specific monitors.

I hope this is helpful. If you want to see more of what ZM is doing to detect motion you can switch on diagnostic images and view each intermediate stage.

Phil

[maciekc]

The motion detection algorithm checks each frame against a reference image. This reference image is a composite of previous frames with each frame blended in at a certain percentage (Reference Image Blend %ge in the Monitor setup). So the reference image at image 'n' is 10% of n-1, 9% of n-2, 8% of n-3 etc. This is to allow the reference image to adapt to gradually changing conditions, e.g. daylight changes to nighttime etc. This percentage is configurable for each monitor, so if you want a more static reference image then use a lower percentage, fi you want a more rapdily adapting image increase it.

You can also modify the behaviour by changing ZM_BLEND_ALARMED_IMAGES in Options->Config. The excerpt from the help explains why,

To detect alarms ZoneMinder compares an image with a reference image which is formed from a composite of the previous images. This option determines whether images that cause events are included in this process. Doing so may increase the precision of the alarmed region but can cause problems if wholescale lighting changes cause alarms as this would not get fed back into the image and an alarm may persist indefinately. A better way to achive the same effect in most cases is to lower substantially the reference blend persentage in specific monitors.

I hope this is helpful. If you want to see more of what ZM is doing to detect motion you can switch on diagnostic images and view each intermediate stage.

Phil

Many thanks to You.
I forget about this option. I had it set to 1 - so it was really to low.
Now I set to 20 and it works perfectly.
Sorry for wasting your time.