Automatic Life Jacket Detection For Drones

garymortimer writes "Sentient, an Australian company that makes drone software, has given UAVs the ability to search for small, high visibility objects such as life jackets. From the article: 'Kestrel Maritime is a software solution that processes electro-optical (EO) and infrared (IR) full motion video (FMV) from manned and unmanned vehicles (UAVs). The Life Jacket Detection enhances Kestrel Maritime EO capability to automatically detect small, high visibility objects whilst searching wide maritime areas.'"

Comments

New Techonlogy?

[Rylz]

This can't be the first software of its kind... Five years ago, NASA was using much more subtle feature detection software in detecting interesting rocks and regions from satellite imagery of Mars. It seems like life jacket detection in oceans would be a much simpler reduction of that problem.

Re:New Techonlogy?

[Eevee]

Well, unless NASA isn't telling us something, there's not a whole lot of ocean on Mars. Spotting a static object without a major time limit isn't exactly the same as having to spot a moving object in choppy seas before the person drowns or freezes, while keeping enough fuel to return to base. (Yes, I watch too much Deadliest Catch.)

Re:New Techonlogy?

[rvw]

This can't be the first software of its kind... Five years ago, NASA was using much more subtle feature detection software in detecting interesting rocks and regions from satellite imagery of Mars. It seems like life jacket detection in oceans would be a much simpler reduction of that problem.

In WWII, they used pigeons to do this. Simply train them to pick a button when they see an orange spot somewhere. Then reward them with some food. I wonder how this compares to the software, in price and performance.

Seems good

[serps]

Nice positioning. Search and rescue staff will still fly around in helicopters with mk1 eyeballs because they don't trust the drone. However, while trying to sell the image-recognition package as a bolt-on to a heli will fail due to pushback from observers, selling in a drone package will be considerably more palatable, psychologically.

Re:Seems good

[camperdave]

Search and Rescue also fly around with pigeons which are trained to peck at a button when they see life-jacket orange. They hit at a 90% rate compared to a human's 38%. Humans in helicopters? If your floating in the drink I think you'd be better off with a drone. Granted, a rescue chopper could pick you up as soon as they saw you, but at nearly a 2/3 chance of being missed?

Re:Seems good

[postbigbang]

No no no. Helicopters are 'last mile'. To run the grid, use drones. XY the 'hits', then nuke^H^H^H^H rescue them.

Re:Seems good

[Daz3d]

Coo.

http://www.uscg.mil/history/articles/PigeonSARProject.asp

Re:Seems good

[Michael Woodhams]

From your link:
"Alvin Wong believes the program would be less expensive today because trainers could use flight simulators to train the birds rather than taking them up in real helicopters."

Only if we could make flight sims with sufficient visual quality to fool a pigeon. Remember, our RGB monitors are adapted to our colour perception, they are not a good reproduction of reality.

Re:Seems good

[oPless]

Use ^W it deletes words ;)

Re:Seems good

[postbigbang]

Old habits are hard to break....

skynet

[soundhack]

Wow, I get that this has search and rescue applications, but the first time I read it I thought this was developed as a way to kill ejected pilots.

Too many flight sim games I guess.

Re:skynet

[e4g4]

search and rescue applications [...] I thought this was developed as a way to kill ejected pilots.

It seems ideally suited to both tasks...

Re:skynet

[im_thatoneguy]

Historically speaking pilots make excellent hostages.

Re:Not just search and rescue

[LordNacho]

What about just having the drones based on land, flying around high-risk areas (possibly only when an alarm is raised), and when someone is found in the water, we send out a helicopter? I don't know about the economics, but my guess is that the drone fleet is reasonable enough in comparison to having a load of helicopters sent out each time a mayday is heard.

Re:Not just search and rescue

[Kinky Bass Junk]

Responding to mayday calls is never a matter of budget. It is always a matter of saving lives as quickly and safely as possible. While scrambling UAVs to search would be good to assist, you would still want to scramble a staffed rescue helicopter at the same time.

Re:Life Jackets

[ISoldat53]

Life rafts and EPRBs on larger vessels tend to deploy automatically when a boat sinks. Having tried to locate an orange life raft from a fast moving jet, I know that they are nearly impossible to see. If this technology can locate something as small as a life jacket many lives could be saved.

Re:Not just search and rescue

[TooMuchToDo]

http://en.wikipedia.org/wiki/Northrop_Grumman_RQ-4_Global_Hawk

"It can survey as much as 40,000 square miles (100,000 square kilometers) of terrain a day."

Performance
Maximum speed: 497.1 mph (800.0 km/h; 432.0 kn)
Cruise speed: 404 mph (351 kn; 650 km/h)
Range: 15,525 mi (13,491 nmi; 24,985 km)
Endurance: 36 hours
Service ceiling: 65,000 ft (19,812 m)

Faster than a helicopter, and with these new algorithms, better at finding people. I say we send the drones.

Re:Not just search and rescue

[jvonk]

...human lives have a maximum dollar amount attached, of course.

Of course they do. Would you honestly argue that it would be practical to spend the entire GDP of the United States to save a single life?

I know this isn't precisely you meant by your sarcastic vituperation of corporatism, but your implication was absolute. If we can agree that "1 life per GDP of US" is above the upper bound of value of human life, then the debate becomes a matter of establishing what the real value is. I read this week in the NYT that US government agencies are currently pegging it in the 4 to 7 million USD range. Naturally, whether this is the appropriate valuation is open for debate.

Re:Life Jackets

[ColdWetDog]

A person just does not last very long in 40 degree water.

I'd be surprised if in the next 20 years it doesn't become common for a cruise ship to have one or two of these just in case.

A good immersion suit can keep you alive for several hours, at least. There are documented survival times of over 24 hours at around 40 degrees F.

Cruise ships are more likely to have UAVs designed to watch for terrorist / pirate activity rather than purely personnel rescue. I've heard of (but not seen) this being actively researched. Having a humanitarian aspect might make them more palatable.