What about combining machine learning and thermal imaging? You would train the ML on the specific thermal characteristics of humans in different conditions. The

What about combining machine learning and thermal imaging? You would train the ML on the specific thermal characteristics of humans in different conditions. The ML may output a ton of false positives, but if there are a lot of eyes on the footage, and other sensors (LIDAR?) + normal image recognition are used, that might be worth it. Another thing to consider is two-way audio search. The drone would make an extremely loud screech to announce its presence to the lost individual (paired with extremely bright light at night), then it listens for a human response. People can be trained to make a certain response that is easier to detect (loud continuous shout? Start and pause every 2 seconds? etc..). Of course, that should all be a backup to safety protocol and satellite communicators. Lastly, I want to make the obvious point that a constellation of drones would do a better job than a single one. A swarm of drones can simultaneously cover the whole area where the lost individual might be. Combined with the above technology, that can be very powerful.