Big Data, Crowdsourcing Aid Search for Flight 370

US-based satellite imagery provider DigitalGlobe is using its recently acquired crowdsourcing website Tomnod to search for missing Malaysian Airlines flight MH370. DigitalGlobe activated its crowdsourcing platform on March 10 in an effort to locate the Boeing 777 jetliner that mysteriously disappeared on Saturday, March 8, 2014, while en route from Malaysia to Beijing.

DigitalGlobe is using its own advanced satellite constellations to take images of the vast search area for the missing airliner. The company uploads the images to its Tomnod platform where volunteers can examine satellite pictures for clues to help locate the missing flight and the 239 people that were on-board.

The best lead so far in the search for the missing aircraft came nearly two weeks after the disappearance. On Thursday DigitalGlobe confirmed that it provided the Australian Maritime Safety Authority (AMSA) with satellite images that appear to show debris from the missing flight. The images were captured on March 16 by the company’s WorldView-2satellite at a resolution of approximately 50 cm. One of the objects was almost 80 feet in length and the other 15 feet. With this possible lead, DigitalGlobe is tasking the satellites to collect more imagery from the area of the Indian Ocean where the possible debris was identified.

The enormity of the search terrain makes this like no other project before. Five high-resolution imaging satellites capture more than 3 million square kilometers of earth imagery each day, a volume that makes real-time analysis impossible, even with a volunteer base that reached two million participants in the first 48 hours. DigitalGlobe has been applying its satellite resources over a broader area than the official search area, while directing Tomnod crowdsourcing volunteers to zero in on the search areas specified by authorities.

“More than 2 million people have tagged some 645,000 features so far, making this the largest Tomnod campaign in history by orders of magnitude,” states DigitalGlobe. “We have continually tasked our satellites to image the ever-widening search area and now have more than 24,000 square kilometers of imagery available for the crowd to comb through. The sheer volume of traffic was a challenge at times for our servers to handle, but we are managing the spikes in activity much better now.”

DigitalGlobe is focusing its search on the oceans around Malaysia. Satellite cameras record thousands of images, which are transmitted into its storage archives. Corrections are made to the photos as needed, to adjust for color disparities and inconsistent camera angles. Then the images are indexed and tagged so they can be searched. The system also automatically eliminates photos that are unusable.

Software vendor Adaptive Computing has been integral to the project. DigitalGlobe uses Adaptive’s Big Workflow package and Moab scheduling software in its datacenters to allocate resources, maximize data throughput and monitor system efficiency to support the analysis of archived Earth imagery. DigitalGlobe’s archives have so far captured over 4.5 billion square kilometers of global coverage. Combining portions of that data with the Tomnod platform, acquired by DigitalGlobe in 2013, allows good Samaritans to assist in search missions.

“The efforts of millions of online volunteers around the world allowed us to rule out broad swaths of ocean with some certainty,” the company said in its statement.

The imaging company’s platform can process geospatial big data in as little as 90 minutes to assist the efforts of first responders and government officials during natural disasters. Adaptive Computing helps DigitalGlobe to operate at a global scale with the timelines their customers need by breaking down silos of isolated resources and increasing maximum workflow capacity.

“When DigitalGlobe gets a call, it’s often under great time pressure to locate an item,” notes Daniel Hardman, Chief Architect at Adaptive Computing. “DigitalGlobe turns their satellites on that particular part of the world to begin the search. They then gather their latest images, and run them through a gauntlet of computational steps utilizing Moab’s Big Workflow solution to aid in accelerating insights. These steps involve taking photo overlays and reconciling the pixels in the photos so they don’t overlap, and then building a giant mosaic of lots of small pictures to determine where an item might be located such as missing Flight 370.”

The effort is still looking for volunteers to support the rescue mission. Interested parties may visit DigitalGlobe’s Tomnod platform to start looking at images.

DigitalGlobe shared in a blog on its website: “We appreciate the work that so many of you have done to search for clues and spread the word, and we sincerely hope the efforts will lead to a breakthrough.”