Guide to Building Data-Driven Organizations in the Public Sector

Eyes in the Sky

Team 5

Topic Overview

Satellite imagery and aerial photography both provide a view of the Earth from above, and both are used to study geography, and to survey areas of land. The methods of creating images differs between the two techniques, as does the application of such images most of the time. While both processes can produce digital images, satellite images have greater large-scale scientific applications, and aerial photography has greater small-scale commercial applications.

Aerial photography is the production of photographic images from balloons, helicopters, airplanes, Unmanned Aerial Vehicle (UAV); it’s used primarily for mapping. Aerial photography is still a better choice for most business and personal commercial uses than satellite imagery. Aerial photography costs less and, in some cases, it’s more up to date, as many available satellite maps are more than a year old and don’t necessarily reflect recent changes or developments. Individuals and small companies can more easily hire an aerial photographer and have more input in the process. Resolution and clarity are likely to be higher as well, making images easier to understand and often eliminating the need for special analysis.

The term “satellite imagery” may refer to several types of digitally transmitted images taken by artificial satellites orbiting the Earth. Satellite imagery has been used for mapping, environmental monitoring, archaeological surveys and weather prediction. Governments, large corporations and educational institutions make the most use of these images. Satellite imagery has several advantages. It can be used to track weather systems, especially dangerous storms like hurricanes, with great accuracy. Satellites circle the Earth, so their imaging activity can be repeated easily. It also allows for much greater areas of coverage and, because all information is digital, it can be easily integrated with software. In some cases, cloud cover does not affect results.

Chapter Summaries

Crowd Computing Aerial Images

This chapter covers a lot of ground from gathering data from satellite and aerial imagery to utilizing volunteers to review the large amounts of data. They go from searching for a flight that disappeared and reviewing data to try to figure out what happened to it to, then onto attempting to locate the tomb of Genghis Khan. They also utilize satellite imagery to figure out how many Somalia refugees there are. With so much data to cover and so little time volunteer resources are utilized to cover so much ground. I found this quote in our reading and it stuck with me “Someone once said that having a map with up-to-date satellite imagery is like having your own helicopter.”

I think one of the things that really amazed me was how with the use of volunteers to help review the data it took just over 2 weeks to tag over a million galaxies. Whereas without the assistance it would have taken well over 2 years to tag all the galaxies with just one person, and that was 12 hours a day for the full week.

Unmanned Aerial Vehicles (UAV’s) come in many different varieties there are fixed wing and rotary-wing variants. Fixed wings UAV’s are like airplanes and rotary wing UAV’s are like helicopters because they can take off vertically and land the same way. UAV’s can be programmed to follow flight paths. Fixed wing UAV’s can fly further and faster, whereas rotary UAV’s need little space to take off, land and can hover over an area. UAV’s can also come in the form of a kite or balloon.

When the BP oil spill happened in April of 2010, often referred to as one of the worst man-made disasters of our time. BP did everything they could to try and restrict access to the area by not allowing boats to get too close, restricting air space and beach access. They did this because they were trying to keep information from getting out on how bad the disaster was. So, Jeffery and other volunteers assisted the fisherman by using a camera attached to a balloon to map out the disaster area. These balloons could be lifted to roughly 800 feet in the air and could capture high resolution pictures.

Artificial Intelligence in the Sky

With the advancement of technology and so much satellite imagery data being collected it becomes difficult to sift through all these pictures using human analysis. So, a European Commission’s Joint Research Center (JRC) created a way to have automated analysis of global high-resolution satellite imagery. Martino Pesaresi of the JRC began with a machine learning, automated classifiers and training data as some of his projects to get this new system up and running. He does this by taking pictures and then uploading them into a database to train his program to recognize when a building has been damaged.

When the tragic earthquake happened in Haiti during 2010 roughly 200,000 buildings had collapsed in the capital and surrounding areas. Martino and his team were tasked with finding the locations of all the collapsed buildings in Port-au-Prince. To accomplish this task, they captured high resolution pictures of the area and began training their system to recognize piles of debris by feeding their system images of the rubble. Once they had utilized their “training data” the result was it took their system less than an hour to identify how much rubble was left and where it was located.

Martino and his team were working on a project to figure out how much growth was happening around the world, but they needed access to a lot of data and the only way to do that was through low resolution pictures from the Landsat satellite. However, Martino and his JRC colleagues needed access to high resolution pictures to have the most accurate results. Martino and his team developed a cross resolution image classifier that figured out a way to take a low-resolution picture and turn it into a high-resolution picture.

Citizen Surveillance

New surveillance technology capable of capturing an area the size of a small city, for several hours at a time (Timberg, 2014) is quickly being developed, however, with it come mixed opinions on the benefits and concerns. Those in favor of this type of surveillance include Ross McNutt, a retired Air Force officer and genial president of Persistent Surveillance Systems. Ross believes that surveillance cameras can help solve crimes that are committed, while also helping reduce crime altogether. A reduction in crime would have positive side effects such as increased property values; as well as lower incarceration rates as the cameras would deter potential crimes. He states that “cameras mounted on fixed-wing aircraft can provide far more useful intelligence than police helicopters do, for less money.”

Those opposed to this type of surveillance are concerned with privacy and personal liberty. The imagery provided from these cameras has limitations. People appear as single pixels making a clear picture of a person’s face impossible to see. Images would aid in tracking movement instead.

In recognition of the concerns, McNutt consulted with the American Civil Liberties Union when writing the privacy policy for his company. Policy includes rules on data retention length, image when and whom can view images, as well language limiting police use occurring only after a crime has been reported (Timberg, 2014). Many police departments are in favor of using cameras to help offset the reduction of personnel. Advances to the technology would increase the range the cameras can capture, not the precision of the image captured.

Key Take-Aways (for Yellowdig)

• Ariel images captured by small UAV’s have advantages over satellite imagery: they fly below the clouds so cloud cover and pollution are not a challenge; satellite imagery is expensive and difficult to acquire; there are licensing restriction on satellite imagery; and UAV’s can take multiple pictures sever times every hour vs satellite’s 1 to 5 day timeframe.

• The Humanitarian UAV Network (UAViators) was created to connect multiple stakeholders to ensure that UAVs are used safely and responsibly during disasters. UAViators facilitate the coordination of UAV flights and encourage the sharing of aerial imagery while setting standards for the use of UAVs in humantiatian settings.

• To efficiently sift through the abundance of information generated after a disaster, classifiers are created to automatically classify tweets based on given criteria. In the case of the Oklahoma tornado, “urgent needs” an “offers for help” were used as the classifiers.

• AIDR – Artificial Intelligence for Disaster Response platform is a free, open source, user friendly platform that was created to allow anyone to create classifiers at any time. This reduces lag time and interpretation delays. Once enough tweets related to the classifiers are tagged by digital humanitarians, AIDR will run automatically, searching for the specified classifiers.

• Collaboration with UNICEF is underway to modify AIDR to automatically classify text messages (SMS) in real time.

• New surveillance technology allows for imagery to be captured across greater areas and for longer lengths of time. This imagery can be used to solve and even reduce crimes. While privacy concerns still remain, poll results indicate that the commercial grade cameras are an affordable alternative and are highly supported by most Americans.

Discussion Questions

• What are the pros and cons with open use of UAVs during disasters, consider the Humanitarian UAV Network?

• Do non-relief related tweets sent during disasters pose an issue in terms of delayed relief efforts? Why or why not?

• Do individual privacy concerns outweigh the benefits of increased public surveillance? Why or why not?

References