Liza Alert: searching missing people using machine vision
habr.com
habr.com
EDIT: Apparently I was wrong.
> Liza Alert takes its name from 5-year-old Liza Fomkina. In 2010 Liza has died of hypothermia in the Russian wilderness after a 9 day unsuccessful search mission. The foundation was born less than 21 days after Liza's death.
I'm actually a member of LA and I know the people who are running this project. So far has never read the article and I just know they're trying to automate image processing with ML.
I'm mostly involved with communications (e.g. radio, etc) and generally helping out with their missions if I can.
They asked if I knew drone pilots awhile back too.
All I can say, get as much sleep as you can, sleep deprivation is a bch.
The article says 22.6% of Russians don't have access to central sewage according to the official data and seven children have been found drowned this way in 2018.
That said I'd expect them to be added at some point. More tools rarely hurt.
There was a case when FLIR had been useful, for example in winter rescue from a drifting ice pack, IIRC.
Also, Liza Alert is cooperating heavily with the 'Angel' volunteer helicopter rescue service. These guys are amazing heli pilots who volunteered to help with SAR missions. My job as a radio operator involves interoperating with them sometimes :)
I'm fairly new and the first time I had to do it I've completely forgotten how to work the pilots, despite being a licensed amateur operator :)
However, even the best thermal cameras are fairly low-resolution -- getting beyond 640x480 generally costs more then a car. So you want to overlay the images, and view them side-by-side, and see if one can tell you something about the other. That workflow could use a lot of improvement.
Also, thermal images have some weird artifacts. In visible light, since we've all grown up perceiving it, shadows and reflections are just natural and we don't even notice them anymore. But medium-wave infrared tends to reflect off things we don't think of as reflective, and it confuses a lot of first-time observers. Different surfaces have different emissivity, and tend to take on the apparent temperature of the background or reflected scene, which is often the sky itself, and the temperature of the sky is a whole 'nother topic. These artifacts are so confusing to untrained observers that they've spawned an entire genre of TV shows ("ghost hunters"), and continually trip up new analysts.
All that being said, yes, adding thermal cameras can definitely produce more useful data. It just might not be suitable for crowdsourcing.