Looking for a needle in a haystack?

It seems the US Air Force needs good ideas on how to identify potential human targets. They're offering a prize of $20,000 for a worthwhile proposal. Wired magazine reports:

The Air Force has problems distinguishing men from women and adults from children. Which means pilots sometimes target - and kill - the wrong people. The air service’s solution: a nationwide contest, to help the military pick out kid from grown-up.

With the "Remote Human Demographic Characterization" challenge, the Air Force is looking for descriptions of a system "that can determine approximate age (adult, teen, child) and gender of small groups of people at a distance". The challenge "requires a written proposal only". So if your idea works and you can get the technical details right, you could walk away with $20,000.

The challenge is one of four that the Air Force Research Lab set up with the launch of its new Open Innovation Initiative. It attempts to use models pioneered by Darpa and NASA to use contests to find solutions to hard problems. Other current Air Force challenges include "Design and Simulation of an Accurate Shooter-Locator" for a whopping $50,000 and "Humanitarian Air Drop" for $20,000.

. . .

Take Monday’s Los Angeles Times story about how mistaken Predator video analysis helped lead to an inadvertent attack on civilians. It’s a veritable play-by-play of how troops on the ground, drone operators, and other intelligence elements - all literally seeing and hearing different views of the same situation - misinterpreted and miscommunicated different pieces of data to misidentify a convoy of civilians as militants.

The biggest failure of the drone operators? Identifying the woman and four children under 6. It’s incidents like this that are causing the controversies in Afghanistan and Pakistan over civilian deaths from drone strikes.

So if the problem is real, why is the Air Force offering only $20,000 to solve it? Probably because right now it’s a problem that is so hard, the Air Force is just hoping for ideas.

Once it has a few that sound plausible it will take it to the Lockheeds and Northrops and Raytheons of the world to build out.

There's more at the link.

I'm honestly not sure that the technology currently exists to get this right. At close range, sure, it's not that hard; but UAV's typically fly high up, where they're unseen and unheard. After all, one doesn't want the enemy to know they're there, and take measures to hide from them (or shoot them down, for that matter). At long range, and in difficult conditions such as darkness, or a dust cloud (e.g. behind a moving vehicle), or surrounded by other people, I'm not sure that the sensors currently deployed can discriminate to the extent the USAF seems to want; nor am I aware of laboratory developments that can do so.

This shows yet again the limitations of "technical intelligence". Sooner or later, the only way to be absolutely sure is to use "human intelligence" to confirm the indications given by technical resources. In a guerrilla warfare environment, that's likely to be rather hard on the human sources. It's not a job for the faint of heart.

Peter