Missing zone crossing...

wpiman

Pulling my weight
Joined
Jul 16, 2018
Messages
332
Reaction score
246
Location
massachusetts
I have a couple of zones on my LPR set up so when a car or plate passes, I get the alert as well as the direction (B-C, C-B etc...)..

This works pretty well. I have an issue occasionally where the trigger happens late. Put in a feature request to have AI test images PRIOR to trigger as well for some programmable number of frames. Ken seemed to like the idea.


The issue I am currently having is a missing trigger. I recorded one below and step through it with the mouse wheel. Camera is set up for 50ms frames. In the video below, I use the mouse wheel to go through each frame. Starting at about 10 seconds in, you see the light from the plate on the left at 30.437...

The next timestamp is at 31.366.... Almost a whole second back....

That has the plate. The next two are pretty much 50 ms split.

I am using maybe 10-20% of the CPU. The AI is done on another machine. Ample memory.

Any idea WHY the video might be missing frames in time? Should I switch back to H.264?


Screenshot 2024-01-29 at 7.38.57 PM.png
Screenshot 2024-01-29 at 7.43.33 PM.png

View attachment Screen Recording 2024-01-29 at 7.37.12 PM.mov
 

Attachments

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,041
Reaction score
48,824
Location
USA
When you posted in your other thread that you did this zone crossing and was getting all the plates, I knew that it would be short-lived (been there, done that).

1706577614887.png

The issue is a tight field of view, an object moving "fast" thru that tight view, coupled with small motion zones, and stuff will be missed.

Really you should use H264, but at a minimum you should use CBR as VBR may not respond fast enough. At night I would watch my LPR bitrate drop into the tens. That is not enough bitrate and to go from 50 to 20,000 can't happen fast enough in that tight field of view.

Less is more. You should have at most Zone A as the whole image and one, maybe two generous sized B and C motion zones.

Something else I noticed is turning on static images (even though it is all black) helped mine tremendously at night.


What are your BI motion settings at night? Since the image is black, you need to make the make time be 0.0, and the min size about a quarter to half the size of a plate in your field of view.


And as I mentioned in response to your yet another thread, I noticed that my LPR camera wouldn't trigger in every instance (I knew this because the overview cam triggered and the LPR camera didn't)

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).
 

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,041
Reaction score
48,824
Location
USA
Here is mine.

Zone A is the whole image and Zone B is this blob in the middle.

The allows cars coming left to right to mostly have the plate in the center of the frame by the time it starts AI. The ones coming right to left then don't trigger until the headlights/front of the car are generally in that area and if the car has a front plate, it gets it.

And then I have it analyze 20 images to account for the different types of vehicles and speeds.

1706578958601.png
 

wpiman

Pulling my weight
Joined
Jul 16, 2018
Messages
332
Reaction score
246
Location
massachusetts
I did have much more luck with fewer zones. The issue is I am trying to determine direction using C>B and B<C.

I did make my zones much fatter. One thing I also notices is you are zeroing out your clock whereas I did not. I watched the motion blocks on an approach and noticed it highlighted the time when the car came in.

I zeroed those out for tonight...

Of course, I am back to bats in the attic so I am totally distracted now by that...
 

wpiman

Pulling my weight
Joined
Jul 16, 2018
Messages
332
Reaction score
246
Location
massachusetts
It is dark out-- but I have been running with this for a few days.

I actually would just like to setup the whole zone to just trigger-- and use another method to determine direction. I'd love it if the AI could do it as opposed to missing a trigger...

Do you know why my capture seemed to be missing frames?

Screenshot 2024-01-29 at 9.17.01 PM.png
 

wpiman

Pulling my weight
Joined
Jul 16, 2018
Messages
332
Reaction score
246
Location
massachusetts
Just to follow up-- with the simple change of masking out the time (and widening up the zones), I have been batting 1000.

In reviewing the footage-- it looks like when an object came into view- the object detection was occasionally highlighting the time as an object and sometimes focusing on that. That object is obviously static and not crossing zones.
 

wittaj

IPCT Contributor
Joined
Apr 28, 2019
Messages
25,041
Reaction score
48,824
Location
USA
Awesome!

Yeah I noticed that early on with some of my cams that sometimes the time changing would get picked up as motion, so whenever I am using zones I exclude the time area.
 
Top