Coral love in the time of Covid-19
The shelter in place period made for one interesting project, live streaming a coral spawning event!
A researcher required a live video feed of up to 6 tanks to remotely monitor active coral spawning. The catch(s): there was no money to do any of this, it had to be done under quarantine procedures, the infrastructure wasn’t in place, the Lab was dimlit/compact, it had to be done ASAP.
Within storage was a PTZ security camera, with onboard h.264 encoding. This meant that it could endure a humid saltwater lab environment, provide views of each tank, and encode video onto the network. A few trips installing hardware and patches allowed for a network connection. One trick however, was that the camera didn’t utilize the youtube protocol of RTMP. This meant that the video would need to be live transcoded.
and this line, or something like it:
ffmpeg -rtsp_transport tcp -i rtsp://password:[email protected]/streamlocation -framerate 30 -video_size 1920x1080 -vcodec libx264 -preset superfast -maxrate 4000k -bufsize 10000k -vf "format=yuv420p" -g 60 -c:a aac -strict -2 -b:a 128k -ar 44100 -f flv rtmp://a.rtmp.youtube.com/live2/xxxx-xxxx-xxxx-xxxx -async 1 -vsync 1
A separate linux server ingested the camera’s RTSP stream, sliced it up with FFMPEG, then directed it over to youtube via RTMP protocol. The camera was then programmed via ONVIF protocols to pan and scan to each tank, and dwell for 20 seconds. This worked surprisingly well, taking advantage of the 16x optical zoom of the camera. One could clearly see in the tanks and it was live on youtube, with no perceptible frame drops.
Of course, after a week, it stopped working.
On top of that, coral spawning occurs just once a year, and that date was rapidly advancing (like within days). Onsite, the camera was found caught in a loop: every time a network service was accessed, it rebooted. Even just http on port 80! This effectively bricked the system, as streaming required a network service. The power supply was swapped, the system was defaulted, and it did nothing. Trying to apply a firmware patch meant using a network service, so that failed too. And not to speak ill of the dead, but the preferred interface of this camera was Internet Explorer.
A small IP67 camera with very low lux was also in storage. Unlike the previous camera, this was not Pan, Tilt or Zoom, nor did it natively encode video. It utilized HD-SDI. Fortunately, a spare SDI streaming encoder appliance was onsite. The PTZ was removed, and the alternate was installed, along with a NEMA box enclosing the encoder, switch + power controller.
The wide view of the fixed view camera was aimed to get the 4 most important tanks, though not a single one perfectly. It was a compromise, but one that worked.
A live broadcast of coral spawning was achieved, with the researcher even doing a live audio Q and A through social media. This last part was done via an RF beltpack combo, straight into the encoder appliance (no DSP to limit noise). This part was was even more last minute, but it was worth it
As always, this took a lot of people to get it working, including the CAS Digital Engagement team, CAS Biologists (working onsite during this pandemic) and the researcher, Dr. Rebecca Albright.
The camera still streams here as of this writing (on an inverted daylight schedule), but this may not be a “forever” install. Or rather, hopefully there will be a more permanent iteration.
2087
stormborns
tangentwithtanner
pixelskylines
snowghoul