[MMM-RTSPStream] - Video Streaming from Live Feeds & Security Cameras

shbatm

@Niggich Just letting you know that I’m testing an alternate method for displaying the video if you want to test it out. It uses omxplayer to overlay the video stream directly onto the display, which uses hardware acceleration. It only works on the local display and is still a work in progress, but there is minimum latency since it’s a direct stream.

To use, make sure you have omxplayer installed and then switch to the development branch:

cd ~/MagicMirror/modules/MMM-RTSPStream
git checkout wip

And update your module config to add: player: 'omx' (outside of the streamX configs).

Future plans include:

Ability to play full screen
Show thumbnails on remote screen when playing locally with OMX
Show ffmpeg stream on remote screen when playing locally with OMX

bhepler

@shbatm Thanks for the different rendering process. It’s still very choppy for me, but it is noticeably better framerate.

Both techniques are maxxing my RasPi’s processor at 100% and I’m getting temperature warnings. But I’m just at the beginning of the process. Let’s see if I can’t cut that down a bit.

shbatm

@bhepler Have you tried the omxplayer method above? This should be using hardware acceleration–on my RPi3 this is using ~10% CPU.

Note: I have updated the development branch to use omxplayer by default for the local screen. You do not need to change anything in your config, just switch to the branch using:

cd ~/MagicMirror/modules/MMM-RTSPStream
git checkout wip
# Update node modules
npm run-script update

See the updated README here for more details – once tested I’ll update the main branch:
https://github.com/shbatm/MMM-RTSPStream/tree/wip

bhepler

@shbatm Whoops. That’s the mistake I made. I switched to the wip branch but didn’t put the omx flag in there.

I’ll try it out tonight. Thanks for the update!

bhepler

Success! I tried the omxplayer version of this module and it works like a charm. The video updates about half a second behind, but that’s well within acceptable limits. The RasPi also is running much cooler and isn’t giving me temperature warnings.

The only problem I have at this point is that the border around the module doesn’t line up with the video. And all things considered, not a big deal. Great work!

shbatm

Great! I’m glad it’s working!

Still trying to find a more robust way for getting the right location for the overlay that doesn’t get messed up by margins and borders.

Starting on line 321 of MMM-RTSPStream.js try changing the +47 at the end of each line. I’d start with making them 0 and see if that helps:

box: {
    top: rect.top + 47,       // Compensate for Margins 
    right: rect.right + 47,   // Compensate for Margins
    bottom: rect.bottom + 47, // Compensate for Margins
    left: rect.left + 47      // Compensate for Margins
}

ICYW: The 47 comes from a combo of the body margin, padding and borders on my setup.

bhepler

@shbatm Again, thanks for all the hard work. Here are my findings (all taken from the development branch):

The video was displaced from the border only on the PI, but not on Chrome nor Firefox whilst viewing the PI’s interface.
Viewing the interface via Chrome or Firefox did show a video image, but it started up FFMPEG and started to heat the Pi up.
Removing the 47 pixel offset made everything line up perfectly on the mirror itself.
Removing the 47 pixel offset made no difference in border position when viewing the Magic Mirror interface via Chrome or Firefox. Everything lined up perfectly.

bhepler

@bhepler Additionally, when I’m running this module, I cannot tell PM2 to restart the MM process. Any attempt to shut down the Magic Mirror and then start it again results in an error claiming that port 8080 is already in use. I have to reboot the Pi in order to get MM running again.
Error message:

Whoops! There was an uncaught exception...
0|mm       | { Error: listen EADDRINUSE 
:::
8080
0|mm       |     at Object.exports._errnoException (util.js:1022:11)
0|mm       |     at exports._exceptionWithHostPort (util.js:1045:20)
0|mm       |     at Server._listen2 (net.js:1262:14)
0|mm       |     at listen (net.js:1298:10)
0|mm       |     at Server.listen (net.js:1376:9)
0|mm       |     at new Server (/home/pi/MagicMirror/js/server.js:26:9)
0|mm       |     at /home/pi/MagicMirror/js/app.js:219:18
0|mm       |     at loadNextModule (/home/pi/MagicMirror/js/app.js:168:5)
0|mm       |     at /home/pi/MagicMirror/js/app.js:163:6
0|mm       |     at Class.loaded (/home/pi/MagicMirror/modules/node_modules/node_helper/index.js:19:3)
0|mm       |   code: 'EADDRINUSE',
0|mm       |   errno: 'EADDRINUSE',
0|mm       |   syscall: 'listen',
0|mm       |   address: '::',
0|mm       |   port: 8080 }

Which is weird, because when I shut down MM, netstat -l doesn’t show anything listening on port 8080. A before-and-after shows the following two processes started (8027 is the port for my IP camera):

tcp        0      0 *:8027                  *:*                     LISTEN
tcp6       0      0 [::]:http-alt           [::]:*                  LISTEN

shbatm

@bhepler Thank you so much for the feedback and the testing.

The video was displaced from the border only on the PI, but not on Chrome nor Firefox whilst viewing the PI’s interface.
Viewing the interface via Chrome or Firefox did show a video image, but it started up FFMPEG and started to heat the Pi up.
Removing the 47 pixel offset made no difference in border position when viewing the Magic Mirror interface via Chrome or Firefox. Everything lined up perfectly.

– This is to be expected. OMXPlayer draws the video on top of the Pi’s screen (over the desktop & electron.js running the mirror). It can only do that on the Pi itself, so if you view it in another browser, the module defaults back to FFMPEG and software decoding.

Removing the 47 pixel offset made everything line up perfectly on the mirror itself.

Good – I will continue looking for a way to improve how the module “finds” the border box to give to OMX. For now this offset may just get moved into a config setting.

when I’m running this module, I cannot tell PM2 to restart the MM process. Any attempt to shut down the Magic Mirror and then start it again results in an error claiming that port 8080 is already in use. I have to reboot the Pi in order to get MM running again.

I noticed this on my system too, but didn’t associate it back to this module. Thanks for pointing it out. Something is not getting cleaned up properly. I am able to restart MM without restarting the whole Pi by killing the electron process manually (e.g. pkill electron or ps aux | grep electron and then kill {pid} on the process number that comes up. I will work on a fix for this in the next update.

Additional Development Notes:

Still working on finding a better solution to using FFMPEG to decode and then re-transmit the stream. With Chrome/Firefox removing support for NPAPI plug-ins, it’s very restricting for RTSP. If anyone has any suggestions, I’m open to them.

Fonfon

Thanks !!! Works great with all my Dahua ip cam
but i don’t know why but i think when my motion sensor put in sleep my mirror my cam disconnect why ?