A Theft, AI Based Perimeter Alarm Triggering and do additional Blue Iris Servers put a load on the cameras?

djborden

Getting the hang of it
Joined
Jul 19, 2017
Messages
50
Reaction score
26
We run two BI servers at each of our locations and one acts as a backup and test server. I am now thinking of adding another 3rd BI server to process AI alerts only for the security system at our office that will act strictly act as an AI based perimeter monitor to actuate strobes and sirens if someone comes within a specific masked areas of our property.

We recently had a trailer theft where the thieves cut our fence drove in our back lot and stole a trailer. Just before the theft, they ran up and banged on a couple doors. I cant help but to think, had I setup AI perimeter monitoring that would have triggered sirens and strobes once they got close to our building and sirens and strobes went off, that would have been enough to drive them away.

I am looking to find out if running additional BI servers pulling from the same camera streams puts additional load on the cameras that may affect performance. At first I didnt think it would as it is just pulling from the same video stream, but then wondered about the process of security and validating each request to access the cameras.

Is there a way to check Dahua camera processor utilization or request latency to check on this or is it a non issue?

David
 

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,158
Reaction score
49,033
Location
USA
Most cameras claim they can have 4 to 12 streams being pulled at the same time (depends on manufacturer).

But of course as you point out, every stream being pulled is a load on the camera processor and at some point it could bottleneck camera capability.

Have you considered instead clone cameras in BI to accomplish the perimeter zones?
 

djborden

Getting the hang of it
Joined
Jul 19, 2017
Messages
50
Reaction score
26
Thank you for your reply! I have experimented with cloned cameras before and it seemed like I would get timing issues with my alerts in the database when doing so when using the app an trying to capture snapshots, or checking alerts on the timeline things were not timed correctly. It could be a coincidence and or that I did something wrong in setting it up cloned cameras.

Do you have any advice on how to properly set that up? I recently discover the "clone master" so have been testing with that as well.

The primary thing I want to do, is use CodeProject AI to reliably and almost instantly trigger a device to sound a siren and strobes. Cloning, or using different zones and alert rules would be ideal so it would not require multiple machines. With that said, I have been getting delays in some alerts and cant figure out why.

I have been testing my process using iPhone push notifications during the day as I am around our property to see how reliable and responsive the alerts have been to determine the viability of this project.
 

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,158
Reaction score
49,033
Location
USA
Yeah I have seen that as well with the clones on some cameras.

What I have done is now that substreams are such a CPU saver, some of my cams I was having trouble with delayed alerts with clones so I came up with a clone hack, which is I simply brought them in as a new camera and used the substream as the mainstream. It uses minimal CPU this way. Since I have the "main" camera doing all the recording, I do not record this clone hack or have it to its own Aux file that only has a few days of retention.
 

djborden

Getting the hang of it
Joined
Jul 19, 2017
Messages
50
Reaction score
26
@wittaj thank you for the tip, I will give that a shot. Maybe that will work for us! You or anyone reading see any issues with the plan and what I am trying to do?
 

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,158
Reaction score
49,033
Location
USA
I think it is a decent plan.

I have something similar.

I have active deterrence cameras on each side of the house that are set up to basically do this.

The cameras are being used as overview cameras, so I did a hack clone and use the IVS rules of the camera to do the strobe/audio, but you could do it with BI AI as well. And a few other cameras can send audio to the active deterrence cameras for the blind spot of the active deterrence camera.
 

djborden

Getting the hang of it
Joined
Jul 19, 2017
Messages
50
Reaction score
26
That is great to hear. I recently ordered some newer Dahua camera with the IVS AI person/vehicle detection. Do you think that would be more reliable with less false triggers than BI with CodeProject AI? I am not familiar enough with the Dahua IVS to know if you can tune the AI to reduce false positives and to only trigger when it is a very high likelyhood of being a person.

I will do some digging but what are you using for the actual electrical relay or switch to turn on the lights or siren?

David
 

bp2008

Staff member
Joined
Mar 10, 2014
Messages
12,680
Reaction score
14,041
Location
USA
Pulling multiple streams does create more load for the camera. Not much, but it is a little and some cameras are already operating right on the edge of their processing capacity.

If you do run into any problems as a result of the added load, your situation may allow the use of rtsp-simple-server to proxy the video streams via a more powerful computer so that less load hits the cameras directly. I use that software for a few of my own cameras such as my wifi doorbell cam or some of the cams I pull 3 or more streams from.
 

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,158
Reaction score
49,033
Location
USA
Many of us use just the Dahua AI in the camera.

 
Top