Can montage be an encoded stream?

[johnyoung]

Hello,
Hello, I'm new to ZoneMinder 1.36.x, I see how to add cameras and how to get individual cameras as a re-encoded stream. But I'm looking to make a quad view of four IP cameras and have that as a single motion jpeg stream. Is that possible with ZoneMinder?

[mikb]

Hello,
Hello, I'm new to ZoneMinder 1.36.x, I see how to add cameras and how to get individual cameras as a re-encoded stream. But I'm looking to make a quad view of four IP cameras and have that as a single motion jpeg stream. Is that possible with ZoneMinder?

I don't know if ZM can do this, but if you have enough computing resources, FFMPEG and LibAVFilter could probably do this outside of ZM. There are filter-graph examples that can join videos side-by-side and overlay them in all sorts of ways. I would think that a quad view would be do-able, possibly with a bit of scaling to keep the resolution under control

e.g. "-filter_complex hstack" at the simplest will take two video sources and put them side by side.

FFMPEG takes multiple input sources, so if you can get it to read your camera's video streams you could be in business. You can chain filters together, so [0:v] [1:v] hstacked in parallel with [2:v] [3:v] hstacked, and then "vstack" them ?

May require a bit of playing with the syntax, building the result one step at a time.

[mikb]

Just tried this for my cameras, using the still image output. They don't provide compliant MJPEG output (it's all framed in HTTP responses that need stripping out, so FFMPEG hates them). However, the principle is the same for video ...

Code: Select all

 /usr/local/src/ffmpeg-3.2.14/ffmpeg -y 
-i http://cam01/image.jpg 
-i http://cam02/image.jpg 
-i http://cam03/image.jpg 
-i http://cam04/image.jpg 
-filter_complex 
"[0][1] hstack [4] ; [2][3] hstack [5] ; [4][5] vstack" -s 640x480 out.jpg

[0] [1] [2] [3] are the incoming video/still from each camera. The filter graph says horizontally stack 0+1, call it new stream 4. At the same time, stack 2+3 and call it 5. And then take stream 4 and 5 and stack them vertically ... the output of which goes into whatever encoding/scaling you set for the video final output.

[burger]

Thanks. I updated the https://wiki.zoneminder.com/Ffmpeg wiki.

I found that the following Ffmpeg wiki link sort of worked for this scenario. Though it did have trouble after a minute or two. I'll have to test it further.
https://trac.ffmpeg.org/wiki/Create%20a ... g%20xstack

[mikb]

Thanks. I updated the https://wiki.zoneminder.com/Ffmpeg wiki.

Note the example I gave wasn't for _video_ as such, as my cameras give not-very-compliant MJPEG-in-a-HTTP-encoded-stream-of-JPEG which is horrid. I need a filter programme which strips the rubbish and passes just the JPEG stream to FFMPEG, but can't figure out how to do that 6 times for 6 cameras

So it only works for still image -> still image, but the principle is probably right.

I didn't know about xstack/ystack=inputs=n, must look into that. hstack/vstack only take two inputs, so doing a 3 by 3 needs lots of hstacking of hstacks, which gets clumsy.

FFMPEG truly is a "Swiss Army Chainsaw" of tools, if you can just figure out the right options, syntax, and order, to get it to do a thing!

[iconnor]

I'm learning about avfilters. I may be able to code this into zms later this summer.

ffpeg is my personal jesus.