Missing very obvious motions

[jeyking]

I am using BI with ten camera's for a few weeks now and have noticed some astonishing missing movements. Even with the motion sensor adjusted at full sensitivity.

For instance:
- A white van driving by, seen at the roof, filling al large portion of the screen, which isn't detected;
- Several recordings of cars suddenly appearing, cars driving up the property, not seeing the person getting out, but some time later do getting in;
- Cars driving by, but just being recorded seconds too late, even with the pre-trigger video buffer set to 5 seconds.

Don't get me wrong; I do like the drive space saving threshold for a motion to be recorded, but I hate it when some so obvious kind of motion isn't being detected. What I've also noticed, is that it's almost impossible to get the motion sensor set to such a high sensitivity that it's nearly recording everything (like moving branches by the wind). Also don't get me wrong with this either - I'm used to software that DID this, just to make sure it was recording wanted small movements - but it's the least you would expect to happen, by setting the sensor motion at the max (100/12/0.0). It just isn't. I'm here viewing a pair of camera's, showing lots of moving branches, temporarily set the motion sensor to the maximum sensitivity, and the're just not being triggered!

To my opinion the part of the algorithm for canceling out unwanted triggers (like moving sunlight through drifting clouds) has too much priority, and for a user it seems to be impossible to set it to a fair compromise. To me, it's unlikely that developpers have overlooked this when testing, which could easily make footage useless for good evidence.

[atreyu]

A hoped for easy fix is to Uncheck “limit decoding unless required” in the camera settings.

With that checked BI only analyzes for motion the key frames coming in. Depending on your camera this can be several seconds. My Reolink cameras have a fixed 0.25 key FPS as I recall. So it only looked every four seconds. Which as you described sucked for motion detection as I also wanted to use BI to trigger Home Assistant to turn on lights. I unchecked it and it’s been great (and the CPU usage jumped from nil to 30%).

Other option to improve triggering after you get detection improved is to run stills through AI to look for things of interest. Then trigger the recording if it sees it (e.g. person, car, umbrella, etc...).

[TimG]

Hi,

I am using BI with ten camera's for a few weeks now and have noticed some astonishing missing movements. Even with the motion sensor adjusted at full sensitivity.

As I have said before, the power of BI5 is that you can adjust everything. A double edged sword, as that also makes it easy to break

1. White van filling large portion of screen - Check Motion Sensor/ object detection/ Reset detector when object size exceeds setting. Mine are set to 85%, but if the van can nearly fill the screen, you may need to increase this value, or even untick it.
2. 100/12/0 - I have the "MAKE time" set at 0.1 seconds. I think setting it to zero may not be a good idea.
3. Pre-trigger video buffer set to 5 seconds - I think this will just eat memory. No problem if you have plenty.
4. What are the pc specs ?

I suggest turning highlight and rectangles on so you can see what is triggering BI and what isn't. Can you show your camera settings and the FPS/Key values ? The key frame is important.

[jeyking]

A hoped for easy fix is to Uncheck “limit decoding unless required” in the camera settings.

With that checked BI only analyzes for motion the key frames coming in. Depending on your camera this can be several seconds. My Reolink cameras have a fixed 0.25 key FPS as I recall. So it only looked every four seconds. Which as you described sucked for motion detection as I also wanted to use BI to trigger Home Assistant to turn on lights. I unchecked it and it’s been great (and the CPU usage jumped from nil to 30%).

Unfortunately this setting isn't checked for al of my camera's.

(The way this forum kicks typed texts to the moon when logged out after a while realy drives me mad!!!)

[jeyking]

Hi,

I am using BI with ten camera's for a few weeks now and have noticed some astonishing missing movements. Even with the motion sensor adjusted at full sensitivity.

As I have said before, the power of BI5 is that you can adjust everything. A double edged sword, as that also makes it easy to break

1. White van filling large portion of screen - Check Motion Sensor/ object detection/ Reset detector when object size exceeds setting. Mine are set to 85%, but if the van can nearly fill the screen, you may need to increase this value, or even untick it.

I will experiment with this
[/quote]

2. 100/12/0 - I have the "MAKE time" set at 0.1 seconds. I think setting it to zero may not be a good idea.

Normally I didn't change it from 1.0, but only to try making the motion sensor max sensitive, I set it to 0.0

3. Pre-trigger video buffer set to 5 seconds - I think this will just eat memory. No problem if you have plenty.

It was indeed and it disappointed me much, cause I was used to set this at 30 sec. or more. But fortunately things got much much better by setting Direct-to-disc recording.

4. What are the pc specs ?

Intel Core i5 @ 3.20GHz 8 GB RAM 64-bit

I suggest turning highlight and rectangles on so you can see what is triggering BI and what isn't. Can you show your camera settings and the FPS/Key values ? The key frame is important.

The key frame rate is 0.50 and drops to 0.25 in night mode (when the frame rate decreased too), but as mentioned above this should not be an issue as long as 'Limit decoding unless required' is not checked, is it?

[atreyu]

As a crude check, how many cameras do you have, what’s the resolution of the scanned for triggers cameras, and what’s your CPU usage? It should be significant (confirmation it’s actually processing the streams), but not pegging out (processing too much bit rate due to cameras * resolution * FPS, and getting behind). Ensure your memory is not over taxed as well and bogging down swapping to disk.

I haven’t played with it, but I’m fairly sure you can take a clip and run it through the trigger settings to see how it should have triggered at different settings. May give some insight as well. And as TimG said, turn on the detect rectangles at least for troubleshooting.

[atreyu]

I reread you have 10 cameras. You are in the range of needing to pay attention to hardware and configuration settings to scan that many. Have you checked the usual guides for optimization?

Detailed guide as well as table of cpu and max bit rate processing
https://ipcamtalk.com/wiki/optimizing-b ... cpu-usage/

Simple big hitters guide
https://www.jesseweb.com/security/optimizing-blue-iris/

My quick list of suggestions to run that many as I suspect hitting either cpu, memory, or writing to disk limit:

• Run Windows and BI on a fast SSD - confirm getting expected speeds, archive to a traditional drive
• Get at least 16, if not more GB ram
• Ensure you’re scanning for triggers on the low resolution feeds
• tweak cameras for FPS and max bit rates

[jeyking]

CPU/RAM is about 30%/1,3 GB, so it runs like a charm.

I haven’t played with it, but I’m fairly sure you can take a clip and run it through the trigger settings to see how it should have triggered at different settings.

What makes you think so? Would be great if it is so, though.

And as TimG said, turn on the detect rectangles at least for troubleshooting.

To me, the rectangles don't realy reveal why one motion causes a trigger and the other is not. However, not only decreasing the min. duration but also the distance an object has to travel surely helped me in getting more motions to be detected.

[atreyu]

How to trial run clips for motion detection: https://blueirissoftware.com/forum/viewtopic.php?t=580

CPU sounds right, but used memory seems low. Maybe I run a poorly configured Win10 but I couldn’t imagine a plain install of win10 alone using that little memory. I’ll check at home, but my work computer has soaked up 10GB for just outlook and some spreadsheets.

True, Rectangles won’t help you know what isn’t triggered, but it does to help know how it is being triggered.

As for TimG’s comment that there are tons of options and easy to get it messed up, you can create a new camera using an existing used video source to “start over” on the settings and see if that helps. The JesseWeb link above does well to describe a basic and optimized camera configuration.

[atreyu]

14GB of 24 memory used. 2.7GB by BI alone. It’s also running home assistant and an AI for image detection. Humph!