Signal to Noise Ratio

One of the most interesting items to me in the Search for Steve Fossett are all of the internet contributors. They are posting coordinates everywhere they can. I am not sure how many are bothering to sign up to Mturk and go through the process.

I think Mturk as a process is very valid but that satellite imagery has limitations. Another major limitation is the amount of noise generated by over eager participants posting every rock they see as a crashed airplane.

Anyway the Main Mturker has been posting a good summary of the scenes that have been examined .
http://s3.amazonaws.com/fossett/index.html
J

A nice Mturk update on the Amazon Web Services Blog

The Amazon Web Services team provides a nice update on the Mturk process on their blog. http://aws.typepad.com/aws/2007/09/update-on-the-s.html
It seems that people have been good at spotting small planes in flight but I have yet to see a crash site.

I also read a nice article in the San Jose Mercury News. http://www.mercurynews.com/breakingnews/ci_6895267
discussing the overall search effort.

I think it would be good if they could provide other types of data, say from the ARCHER system that CAP has been using which has hyperspectral imagery. I wouldn't provide this to the masses but qualified persons which is something you can do in Mturk.
-J

It is hard to spot crashes in satellite imagery

I wanted to see what an actual wreck site looks like in sat imagery. This is a Learjet 35A that crashed in New Hampshire in 1996.

If you zoom in on the coordinates you can see how hard it is to spot imagery. Look here on Google Maps



View Larger Map.



I am still looking for a good crash site example in the western US. Any ideas?
-J

Summary of the Imagery analysis on MTurk

Wired has a nice summary of the imagery analysis so far. If anything this process shows how difficult it is to find a small object such as a light aircraft in satellite imagery.

The search effort would probably benefit from using video captured from small planes. This effort has demonstrated the ability to examine a huge amount of data

Read the whole article here but I clipped some highlights below:

http://www.wired.com/software/webservices/news/2007/09/distributed_search

*Up to 50,000 people are hunting for missing aviator Steve Fossett, in a distributed search using satellite images and Amazon's Mechanical Turk.

*A 17,000-square-mile region search area.

The "distributed search" uses satellite images from DigitalGlobe, the company that provides images for Google Earth.

*each image is a 278-foot-square pieces of the search area.

*One of the heartiest of participants is 25-year-old software designer Andy Chantrill of Castle Donington, England, who worked on it for 13 straight hours on Monday and says he's put in 30 total hours examining about 5,000 squares.

*Amazon was happy to release the number of satellite images that have been scrutinized: more than 2 million, according to the company.

*There are more than 300,000 squares in the map. Each square is viewed 10 times, and the entire map has been re-entered at least three times as newer satellite pictures of the area have become available.

Help Search for Steve Fosset

If you follow search and rescue in the news you might know that Steve Fosset, billionaire adventurer is missing in the Neveda Desert.

You can now actively participate in the search.

Go to http://www.mturk.com/ and get registered then search for the "Steve Fossett Missing: Help find him by searching satellite imagery" HIT.

You need to have Google Earth installed on your computer. Follow the directions. I wont repeat them here because they are detailed nicely on Mechanical Turk.

This approach was used in searching for the missing sailor and famous Microsoft Computer Scientist Jim Gray earlier this year.

I will talk about the technology later but go ahead and get started!!
-Jim