Troubling shooting my motion detection.....

W

wpiman

Getting comfortable
Jul 16, 2018
378
264
massachusetts
So I have an ALPR camera that watches my street and I am happy with the photos it provides. Both day and night. I am also pretty pleased with the ALPR function, and I run codeproject AI on another machine and that runs well.

iframe rate 24. frame rate 24. I shut off substreams for the time being to reduce variables. I noticed my substream was 20 and 20. I switched it to 24 and 24 and might try that again later.

I have it set up for zone crossing so I can ascertain direction. Zone B is on the left. C is on the right. I trigger on an object moving B-C. Birdirecitonal. Very quiet street so motion detection SHOULD be simple.

I set the sensitivity pretty low, and that doesn't seem to matter. All the way down didn't help. I am wondering if maybe I set it too low.

I have some cars go by and it just plain misses them. I recently set up a clone of this camera and just set that to simple motion and that clone is doing much better, albeit not perfect.

I posted a video of a big ass car in broad daylight that didn't trigger on zone crossing motion, but did trigger on simple.

I was starting to suspect that maybe I need a faster machine. I am on an E3 1245 which does have Quicksync acceleration, but it is old. I like it because it has ECC memory and it is stable as a rock. When I get a trigger, I noticed the CPU went from 20% to 100% for an instant.
However, a huge ass plow just went by neither the simple or zone crossing moved and my CPU stayed pegged at 20%. The machine only runs BI and BI tools.

Am I missing something basic?


Screenshot 2025-02-13 at 7.57.06 AM.png


Screenshot 2025-02-13 at 7.57.40 AM.png

Screenshot 2025-02-13 at 7.56.50 AM.png


View attachment Clipboard_LPRNOZONE.20250213_074153-074157.24.mp4

Screenshot 2025-02-13 at 7.56.29 AM.png
 
I had a similar issue.

So I noticed while watching it live for an extended period that for my license plate camera (which as you know is zoomed in tight to the road to read plates), I watched it not trigger for a big ole yellow school bus, but then trigger for a tiny 2-door car the next minute that was driving slower and then miss the same car coming back 5 minutes later!

For this plate camera, I was obviously running a fast shutter to capture plates, but also had the FPS at 30 FPS thinking that would be better. When I knocked it down to 10 FPS, Blue Iris motion started capturing that bus and other vehicles it was missing and triggering faster.

I think the motion algorithm for a tight field of view was having difficulty with the faster FPS as there wasn't as much of a difference comparing frame to frame at 30FPS to 10FPS. A vehicle is in and out of my LPR field of view in under 0.5 seconds and I now get trigger alerts and capture every plate at 8FPS (yes I dropped it even further for longer retention of LPR images).

So lower your FPS and see if detection improves.
 
  • Like
Reactions: Cooltiger and wpiman
wittaj said:
I had a similar issue.

So I noticed while watching it live for an extended period that for my license plate camera (which as you know is zoomed in tight to the road to read plates), I watched it not trigger for a big ole yellow school bus, but then trigger for a tiny 2-door car the next minute that was driving slower and then miss the same car coming back 5 minutes later!

For this plate camera, I was obviously running a fast shutter to capture plates, but also had the FPS at 30 FPS thinking that would be better. When I knocked it down to 10 FPS, Blue Iris motion started capturing that bus and other vehicles it was missing and triggering faster.

I think the motion algorithm for a tight field of view was having difficulty with the faster FPS as there wasn't as much of a difference comparing frame to frame at 30FPS to 10FPS. A vehicle is in and out of my LPR field of view in under 0.5 seconds and I now get trigger alerts and capture every plate at 8FPS (yes I dropped it even further for longer retention of LPR images).

So lower your FPS and see if detection improves.
Click to expand...
Ok, I set it for 10 FPS across the board.

Are you using substreams? I set those for 10 too...
 
Make the min contrast the minimum and have the two zones touch without a gap.

And zone A is the entire screen?
 
I made Zone A the left half of the screen and Zone B is right up against it on the right hand side. Set it to the most sensitive motion and changed from B-C to A-B. No gaps. (one of the videos on line recommended them)

Changed the FPS to 10....

I also looked at my videos and realized now that my trigger happens late and I sometimes miss the plate. I upped my AI pre-trigger images to 15 (1.5 seconds) and post trigger to 10 (1 second). I've only missed on plate on the Motion crossing this morning... I've gather some more data but it is definitely looking much better.

Screenshot 2025-02-14 at 9.38.13 AM.pngScreenshot 2025-02-14 at 9.42.56 AM.png
 
wpiman said:
Ok- so with about an hours of traffic-- the simple algorithm definitely saw improvement going to 10 FPS.

The zone crossing still missed quite a bit.... This car went by--- totally missed...

View attachment 214416
Click to expand...


Not sure it applies the same to BI (I'm not a BI user) but on Dahua NVR's using AI, I've noticed that vehicles can be missed like this one when they exceed the size of the screen.

I'll occasionally see a big garbage truck or other that exceeds the screen of a zoomed in camera. Bu zooming out just enough to allow the vehicle to be within the screen it catches it.

Maybe nothing but came to mind when i saw this clip
 
You must log in or register to reply here.
Top Bottom