Analyze sub-stream, Record Main stream

[jbmia]

Hi guys,

Haven't posted in quite a while... started with 1.24, went to 1.25 virtual machine, and recently installed 1.26.5.. I'm currently running running an 8 core / 16 gb Centos server with zm running 4 cores & 4 gb's in an Ubuntu KVM guest. New version seems to run much smoother than older ones... No on to the topic:

I recently installed a Dahua IPC-HFW2100 720p ip camera. I'm considering adding more of these or even going with their 1080p version. Right now, I'm seeing maybe 10-15% of a core for zma & zmc for this camera... My older 640x480 cams run much less. If add a 1080p cam, I could see the core usage go up much higher.. though I haven't done any testing or research. But this got me thinking... all these newer ip cams seem to run a main stream (e.g., 1080p) and a sub-stream at a lower resolution (e.g., D1)... My question is would it make sense to to use both streams? The sub-stream for analysis and record the actual event based on the main stream? It would seem there would be some processing savings here at the expense of some bandwidth maybe (two streams).

Can we do this today? Is this something that would make sense moving forward as camera resolutions increase? Or.. should we just grow our processing power as we grow our resolution??

Any thoughts?

[mikb]

My question is would it make sense to to use both streams? The sub-stream for analysis and record the actual event based on the main stream?

Not tried this, but, I believe "linked cameras" is what you're looking for. Something along the lines of :-

Create "monitor" 1a, point it to your 1st camera's low res stream, configure all the motion detection, and set it to Modetect
Create "monitor" 1b, point it to same camera's high res stream, and link it so that it records only whenever 1a records.

Repeat these pairs for each camera.

This should get you heading in the right direction.

[jbmia]

Thanks for this. The only minor issues I could see with this approach would be that I'm actually capturing two events.. one on the sub-stream monitor and another on the main-stream monitor. And, any image outlining would be done on the sub-stream.. Thanks though.. decent work around.

[mikb]

There may be a way to kill off the "false" events with a filter (purge with extreme prejudice anything on the unwanted camera older than X mins/hours).