Cancel
Advanced Search
KEYWORDS
CATEGORY
AUTHOR
MONTH

Please click here to take a brief survey

Rapid Image Evaluation for Disaster Response
David Zaks, 6 Feb 07
Article Photo

Access to information is critical in the early hours in any post-catastrophe scenario. Quickly relaying information to emergency personnel can reduce the time from disaster to response. When calamity strikes over a large area, the constellation of satellites overhead are crucial in the primary assessment of the scale of the problem, and in assigning resources to the most appropriate places. Recent disasters have shown the effectiveness of satellite imagery, and the power of the Web to mount an effective response. These tools are currently being put to use in the search for Jim Gray, an acclaimed computer scientist at Microsoft and a competent sailor, who disappeared last week after a trip to the Farallon Islands to scatter the ashes of his late mother. After not returning, the Coast Guard was notified and conducted searches for four days; however, after finding no trace of the boat, they called off their efforts.

But the friends and colleagues of Jim Gray were not ready to give up. Leveraging nearby Silicon Valley's collective technological resources, a concerted response by Amazon, Google, Microsoft and NASA has permitted a continued search. Teams were created to initiate searches, harnessing expertise from areas like distributed computing and image processing. DigitalGlobe's Quickbird satellite was tasked to acquire imagery of the region, but though their help was timely, they faced scattered clouds which could obscure the boat from view. Their high resolution imagery was complemented by imagery acquired by a NASA ER-2 aircraft aircraft which was already scheduled to complete a training mission, and so agreed to capture imagery to aid the search.

With the tremendous amount of data acquired and processed, Amazon engineers working on the project developed a method to split the images into manageable pieces, then upload them to the Mechanical Turk that is set up for repetitive tasks that require people -- or as they say, "artificial artificial intelligence." With a bit of training in how to differentiate clouds, land, open ocean and objects like boats, anyone can log on and scour the imagery for signs of the missing boat, and indeed they have, with over 12,000 people inspecting 530,000+ images. Each page contains several images and radio buttons where users note whether the image should be forwarded to an expert for further review, or contains nothing of interest.

While this software package was assembled relatively quickly in an emergency situation, the foundation has been laid for a powerful collaborative disaster response tool, and can be used with the available satellite imagery for object identification, as in this case, or change detection where post-disaster imagery could be compared to older imagery to look for potentially catastrophic changes. While algorithms have been developed to sift through this type of imagery, sometimes the human eye is vastly superior.

As concerned citizens across the world sprang to action in light of recent disasters, the search for Jim Gray has added another tool to the information toolbox. While we all wish that systems like this could be developed before actually needing to be implemented, the speed at which the system was put in place and the groundswell of support it has received show how goodwill can lead to powerful collaborations and faster disaster response.

Click here to help find Jim Gray.


(Thanks, Sam!)

[image]

Bookmark and Share




EMAIL THIS ENTRY TO:

YOUR EMAIL ADDRESS:


MESSAGE (optional):


Search Worldchanging

Worldchanging Newsletter Get good news for a change —
Click here to sign up!


Worldchanging2.0


Website Design by Eben Design | Logo Design by Egg Hosting | Hosted by Amazon AWS | Problems with the site? Send email to tech /at/ worldchanging.com
©2012
Architecture for Humanity - all rights reserved except where otherwise indicated.

Find_us_on_facebook_badge.gif twitter-logo.jpg