Seeking some guidance choosing first home camera setup

[isabido]

Hi Wittaj, This is how it looks right now with these light conditions, today is a cloudy day, but normally there is a lot of sun. It is very likely that the image will end up "burning". I see this type of "FIXED" configuration as correct in night mode where the light conditions are almost always the same, but during the day it is too changeable. Don't you think?

 

[isabido]

It is incomprehensible!! same /same lighting my daughter, who is smaller, detects it as soon as she enters and it doesn't detect me until I'm halfway down the stairs.

I'm gonna go crazy!

 

[isabido]

In this other capture, my daughter and her friends of similar size, and now it runs later too. It is something unpredictable and non-repetitive, so it is difficult to correct. I'm bored of drawing lines in all possible ways.

 

[wittaj]

It appears to be a contrast issue.

The first clip it appears your daughter is coming from the dark grass so there is enough contrast. You come up the light tile with a white shirt so not enough contrast.

Then the 3 girls it goes red on the girl with dark clothing against the light tile.

Have you tried the global configuration to calibrate? You also have the camera seeing two different elevations that can be contributing.

And you might need to just add a zigzag trip wire along that door as well. If that door is always closed, I would consider a single trip wire in a zig zag and turn off object detection to let it trigger on the door opening?

 

[isabido]

Thanks Wittaj for all the interest in making this work.

I have doubts regarding the calibration procedure. But I think this has not caused any improvement.

I think that one of the problems may be the distortion of the wide angle. and trying to detect in "depth / 3D" at very close ranges.



About the option of using a zigzag tripware on the door only it is not possible, the normal thing is that if someone comes to steal they will not open the door if they do not jump over the wall/fence. Hence my intention to cover the perimeter with IVS rules.

I'm going to try this ugly drawing... I don't like it much!

 

[wittaj]

Did you then do the 4 measurements with the calibration?

Who cares what the IVS rules look like as long as they work!

 

[isabido]

The horizontal verification did not get it to work, right at the calibration point it has a 1 meter.

If they have 3 vertical lines and 1 horizontal.

The horizontal verification did not work, just at the calibration point it has a 1 meter, but when I move it to one of the corners of the image the measurement is totally illogical (36.23 meter)

Let's see if you can test

 

[wittaj]

Yeah something isn't right there. If you are using 1 meter as the test, it should come back plus or minus 20% or so.

I would try it again until you get it to register it correctly.

It is a tough field of view, but we should be able to figure out the right combination of IVS rules or SMD to accomplish what you want.

I am thinking you need to add one IVS rule that doesn't use object detection and make a min size about the size of your daughter so that a bird won't send it off, but have that rule as a backup if the human detection doesn't catch it?

 

[isabido]

The SMD human detection works perfectly, but in this camera model Dahua has not felt like enabling the light/sound trigger.

Regarding the suggestions of minimum/maximum size and filter of objects I have also tried with it.

Although I have to say that I have never been able to trigger an IVS rule with any object that is not a HUMAN or VEHICLE. Is this correct? should I trigger with a ball for example? Do you remember when I was testing with the pillow/cushion? it was not able to detect with the object filter disabled and a very very small minimum size.

About the calibration I have tried in several possible ways and I can't get it to work correctly, I don't know if it's the perspective, angle of view of the camera or the calibration "box".

 

[isabido]

I seem to be getting worse instead of better!
View attachment Playback - Internet Explorer 2022-09-01 17-21-42.mp4

 

[wittaj]

It is a head scratcher.

Try one big zigzag trip wire that starts at the top of the wall and zigsags from left to right.

 

[isabido]

It's more effective, of course. But not perfect as I would like.




Here you can see how effective it is with side scrolling from left to right, it works perfectly.

But it works really badly and inaccurately, with the movements from back to front in the axis of the camera.

I am sure that if they were performed at a greater distance, they would work better.

The system needs to have a depth reference to do this correctly, I understand that it could be solved with the calibration, but I am not able to.

Wittaj, you could do some tests on your system to see what results you get.
View attachment Live - Internet Explorer 2022-09-01 19-02-55.mp4
View attachment Live - Internet Explorer 2022-09-01 18-56-20.mp4

 

[wittaj]

I was thinking more something like this but tighter:

 

[isabido]

With a smaller intrusion area, it seems to work better, the problem is evident in depth of field, for an intrusion box to work well it needs not to be too close and of a size proportional to the view.
View attachment Live - Internet Explorer 2022-09-01 19-29-40.mp4

 

[wittaj]

Yeah the reason it works better with a smaller box is the camera needs time to see the object, determine if it meets the criteria, and then trigger if it does. But some field of views that is tough to do.

But so many factors come into play as well - the field of view and setup of the camera. The FPS and bitrate impacts it. At the end of the day, these basically have a little processor in them and if we overload them trying to run max settings and every function available, then when it pegs at 100% something gives and it is usually something like recording or triggering.

 

[isabido]

I totally agree with your contributions. Finally in my scene, it would be enough for me with SMD, I could even disable smart plan IVS. And unload the processor. But damn it!! Dahua did not activate the light/audio trigger in this firmware!! Who could I cry to? Will Dahua listen to this request? or is it because they want to limit it only to Tioc cameras? In addition, as I have been able to find out from the images and videos of other users, this function is only enabled in TIOC 2.0

 

[Sammyf]

For the global setup I found on some Dahua`s manual that the calibration region should be on same horizontal plane. So in your case for example on the floor downstairs.
When making the vertical measurements, the bottom of each measurement should be on that same level (on the floor downstairs). Height of the physical object doesn’t matter, you could use one object and move it around.

With your field of view it might be difficult to fit high object inside the calibration region. The measurements should be made in triangular form. All these things were for me useful detailed info to supplement the info found on dahuawiki.

 

[isabido]

Para la configuración global, encontré en algunos manuales de Dahua que la región de calibración debe estar en el mismo plano horizontal. Así en tu caso por ejemplo en el piso de abajo.
Al realizar las medidas verticales, la parte inferior de cada medida debe estar en ese mismo nivel (en el piso de abajo). La altura del objeto físico no importa, puedes usar un objeto y moverlo.

Con su campo de visión, puede ser difícil colocar objetos altos dentro de la región de calibración. Las medidas deben hacerse en forma triangular. Todas estas cosas fueron para mí información detallada útil para complementar la información que se encuentra en dahuawiki.
[/COTIZAR]

Excuse my bad English from Spain, it has been difficult for me to understand you. Could you share that information link or paint the idea on onemy scene. Thanks in advanced!

 

[Sammyf]

Unfortunately I don’t have now access to pc to draw on your picture (and too clumsy to do it on mobile device), but here is a link to same info I found earlier:

https://mans.io/files/viewer/2153084/142#navigate_bar

So in the instructions the horizontal plane is the street, for you maybe the floor downstairs.

And the example tells to use for example light poles for vertical measurements, you could use for example a chair or even side of a cardboard box. You would put the box on the ground, draw ruler to match the height of the box, move the box, draw ruler again and do it for third time.

For horizontal measurement the example suggests road sign, you could put for example a broom laying on the ground and use that for measurement. Or you could use the same box and just draw the “depth“ (or width) of the box, measured on the part that is against the floor.

It is a bit hard to explain the process, but hopefully this helps you. But it also seems that your field of view might be challenging for making the calibration.
For two of my front door cameras installed at approximately face height, set in corridor mode and people approaching almost directly towards (and away from) the camera I could not get the IVS working perfectly even with the calibration. Maybe still playing with image adjustments is needed.

However, cameras are addictive (lol), so I got additional overview camera that will be installed to also work as a spotter for these cameras. The overview camera will be used to observe/recognize and it will trigger via Blue Iris the front door cameras used for identifying. Also, more viewing angles is better. Of course all this comes with added expense of getting the overview camera, but for security purposes I think it is worth it and buying cameras over time makes it financially doable.

 

[isabido]

I have already managed to understand the procedure and it seems that I have done it correctly. The correct thing is to make the ruler horizontal in the depth axis (can we call it Z axis? ), as it does in the documentation and as you can see in my image.

Now it seems with the verification ruler it is able to tell me the measurements on the Z axis.






floor width
parking width

road width