Cognitive Daily

How to search a large area

When someone gets lost in the woods — or when a convict escapes from prison — finding them among the vast expanse of trees and other features can be a daunting task. Often search parties literally walk in lines just a few feet apart to scour the terrain for evidence. But perhaps there’s an easier way to find signs of life. When cNet editor James Kim and his family went missing last week somewhere in Oregon, Matt Haughey had an idea about how to find them:

I’ve seen some pretty amazing stuff come out of MetaFilter, when people collaborate on a real world problem. Then it hit me. There are only 5 or 6 major roads to the coast, and they’re not that long. Why not run a surveillance plane 500′-1000′ feet above each of the roads, going slow enough that it takes maybe 20-30 minutes to follow the roads to the ocean. If the camera view could capture 100-200′ north and south of the road, you could probably film all 5 or 6 major roads in a single clear day like today.

If each recording is say 30 minutes long for a road, split it into 10 equal parts, 3 minutes long, and upload all of them to youtube. Ask viewers to leave comments pointing out when they see anything strange. The Kims were in a silver Saab wagon, so it’s probably something that can be seen from above. In total, there’d be 50 or 60 short clips and in a matter of hours you could have millions of people closely scan then and start pointing out the things worth looking into on the ground. If everyone says there’s a silver glint in the trees on video #6 from the highway 18 group, at 1:55 in, you could send a police unit out to investigate.

Kim’s wife and children were located, safe but snowbound in an isolated Oregon canyon, by a helicopter using thermal imaging equipment. Kim is still missing. Perhaps with thousands of eyes going over the data from aerial surveillance, he could have been found.

A separate but related problem: how to monitor the same forest for signs of a forest fire. A report in ScienceDaily covers one approach:

A forest ranger helicopter flies over a forest, scattering sensors that can relay temperature data to the ranger station. To ensure minimal environmental impact with maximum robustness, the sensors are very simple: they are basically tiny, sturdy thermometers. After the sensors are scattered, they might be moved further by winds, rains, rivers, or even animals. Is there a way to take the local information sent by the sensor network and turn it into global information about the existence and location of fires in the forest? In particular, without knowing the exact locations of the sensors, can one nevertheless glean information about the coverage area of the sensor network?

It turns out, using mathematical modeling, they may be able to. This technology, too, might be applied in a search for missing persons.

In other news:

Comments

  1. #1 Mark
    December 6, 2006

    I guess it’s related, but only slightly so, but I noticed something the other day while walking on an almost-never-used road covered with leaves. Someone had driven a car on the road a day or so earlier and I was trying to follow the tracks to see how far they went. I noticed that I could see signs of the tracks while quickly scanning up the road, but that when I tried to examine a small area closely I could hardly see any sign. That indicates to me that there is some pattern recognition going on that can identify subtle differences in relatively large areas, and that those differences can be so slight as to be essentially hidden by noise when examined in detail. That seems to be an indication that a quick, aerial scan might be useful when searching for lost people, even aside from getting lots of eyes to look at the results.

  2. #2 Sean Thompson
    December 6, 2006

    In response to the above comment.
    I agree with this ‘can’t see the forest for the tree’ idea of focused attention missing the detail. Animal trackers apparantly relax and stop ‘trying’ to look for signs of an animal, so that their senses are more open and they get a feel of the general ambience around them, any disturbance to this status quo then ‘jumps out’ to them.
    I try this approach whenever I lose a pen / mug or something on my desk , it doesn’t always help!

  3. #3 natural cynic
    December 6, 2006

    That indicates to me that there is some pattern recognition going on that can identify subtle differences in relatively large areas, and that those differences can be so slight as to be essentially hidden by noise when examined in detail.

    This is precisely what happens when you look at an impressionist or pointillist painting. When I focus on a small area and the fine detail of a Monet, I see something almost like Jackson Pollack painting – almost nothing is recognizable – it’s all points and splotches of color. Moving farther away brings the easily recognizable figures into focus.

  4. #4 anomalous4
    December 8, 2006

    Tragically, Mr. Kim was found frozen to death, day before yesterday, I think.

    The YouTube idea sounds great. If you put out a short announcement on the evening news and the online news services, YouTube would get bombed with hits. My only concern is that their server(s) might get overloaded. That wouldn’t help matters at all.