Satellite Imagery Comes Down to Earth

Hank Hogan

It used to be that, to get a bird’s-eye view of your house, you had to be a bird — or somebody in the intelligence community. That’s not true anymore. The advent of software from Google Inc. of Mountain View, Calif., Microsoft Corp. of Redmond, Wash., and Yahoo! Inc. of Sunnyvale, Calif., as well as a planned relaunch of an application from AOL’s subsidiary Mapquest Inc. of Denver has brought satellite and airborne imagery down to Earth. As a result, people can fly, in a virtual sense, over buildings and down canyons. These trips can be over familiar places, or over exotic or previously hidden locales. And they don’t require extraordinary knowledge or expertise.

“Anybody with a computer can go on and see any location on the planet. We are entering an age of transparency,” said Mark Brender, vice president of communications and marketing for satellite imagery company GeoEye Inc. of Dulles, Va.

Satellites such as GeoEye Inc.’s Ikonos, shown here, can take high-resolution images of Earth.

Behind these views from above lie advances in imaging, improvements in software and an ongoing desire to make a profit.

Advances in imaging come from improvements in photonics technology from such companies as GeoEye, DigitalGlobe Inc. of Longmont, Colo., and Spot Image SA of Toulouse, France. DigitalGlobe is one of Google’s imagery suppliers, and GeoEye supplies imagery to Yahoo! and Microsoft.

GeoEye Inc. operates a 1-m, high-resolution satellite named OrbView-3 (top). Another satellite, called GeoEye-1, will be launched early next year (bottom).

Brender noted that GeoEye operates two Earth-imaging satellites, Ikonos and OrbView-3. Both fly in polar orbits at an altitude of about 420 miles, passing overhead from pole to pole at about the same time every day. As the Earth rotates, points pass below the satellites and are imaged with both a 1-m-resolution panchromatic camera and a 4-m-resolution multispectral device that captures blue, green, red and infrared. In the case of Ikonos, the data can be combined to yield a 1-m-resolution color image. The offering from DigitalGlobe has somewhat higher resolution, while that of Spot is lower.

Image acquisition

One advantage of the polar orbit is that the images are consistent over time. “As you look at one part of the world, you’re looking at the same resolution, the same shadowing, the same quality imagery,” Brender explained.

GeoEye downloads collected data, stores it in a raw format and then sells it to buyers after it has been processed to their specifications. As of June, the company had imaged about 250 million square kilometers of the Earth’s surface.

When a particular spot has been imaged, there is some delay before the data is available. Thus, although the images are collected in real time, what is viewed comes out of archives and may not be current. Google, for instance, has stated that the images presented by its viewer are up to 3 years old.

A technician inspects the silver-colored payload adapter of Ikonos prior to its Sept. 24, 1999, launch. The large solar arrays are in a stowed position awaiting encapsulation inside the launch flaring. The circular black star tracker and white GPS sensor are clearly visible. The star tracker helps determine the orientation of the satellite by monitoring the stars, and the GPS sensor determines the satellite’s position relative to the latitude and longitude on Earth.

DigitalGlobe plans to launch a satellite with higher resolution within the next few years. GeoEye also has a new satellite, GeoEye-1, scheduled for launch early next year. It will produce a ground resolution of 0.41 m panchromatic and 1.65 m multispectral. Merging the two in postprocessing on the ground should yield 0.41-m-resolution color imagery late next year, making it the best commercially available.

At resolutions better than that, satellite imagery encounters various problems, some technical in nature. For example, it is possible to image at 0.25-m resolution. However, the resulting data sets would be large and difficult to download, store and manipulate.

Another problem is economic. Satellites capable of high-resolution imaging would be large and expensive, and imagery suppliers might have a hard time recouping the extra expense.

Finally, there are regulatory challenges. Satellite imagery companies operate under licenses from various government agencies, such as the National Oceanic and Atmospheric Administration (NOAA) in the US.

Adrian Zevenbergen, managing director of Munich, Germany-based satellite imagery supplier European Space Imaging GmbH, noted that this may be the most important consideration of all. “The limiting factor of resolution for commercially available imagery is set by the US government — NOAA — not by technology,” he said. “It is widely acknowledged that the US government has systems with a much better resolution.”

Such companies as Google and Microsoft supplement satellite images with airborne ones. Microsoft has partnered with several aerial mapping companies, including Earth-Data of Frederick, Md., to provide this data. In May, the company bought Vexcel Corp. of Boulder, Colo., which makes remote sensing systems and software with a specialty in photogrammetry to determine an object’s three-dimensional coordinates from two or more images.

Aerial shots, such as this one taken by the UltraCam, are used to supplement satellite imagery in virtual mapping applications. These shots provide details and resolution that can’t be achieved otherwise. Three-dimensional information can be extracted using overlapping images taken from various angles. Courtesy of Vexcel Corp., a Microsoft company.

One of the things that Vexcel brings to the deal is sensor technology, such as its flagship Ultra-Cam digital aerial mapping camera with 13 CCDs. Nine of the sensors are used to create a large-format panchromatic focal plane, while the other four provide red, green, blue and infrared channels.

Imaging parameters

William B. Gail, who was with Vexcel before the acquisition and is now a director within Microsoft’s Virtual Earth business unit, noted that the company’s first choice would have been to build the camera with a single 14-bit CCD for the panchromatic focal plane, but that that would have called for a sensor with roughly 14,000 3 9000 pixels. Because such chips cannot be fabricated economically with today’s semiconductor technology, the company chose CCDs of a more reasonable size and price, combining multiple images to create one panoramic whole.

The UltraCam has 13 CCDs. Nine of the sensors create a large-format panchromatic focal plane, and the others provide red, green, blue and infrared channels.

This software merging is helped by the sensor layout. Four CCDs are mechanically and rigidly connected, forming the four reference corners of the image frame. The remaining five are arranged in horizontal or vertical pairs, with one in the center. The nine images are stitched together via software into a single image. The sensor arrangement ensures that errors associated with the process aren’t magnified because parameters needed for the merging are derived from interpolation and not by extrapolation.

According to Gail, another consequence of this approach is a faster readout than would be possible with a smaller number of CCDs. That boosts the frame rate. That extra speed can provide redundant information that is very useful when automating photogrammetric software, he said.

Another key driver for such automation is an improved signal-to-noise ratio in the digital images as compared with film imagery. Better CCD chips, carefully designed electronics and more powerful software algorithms play a role in achieving that ratio.

Gail said that the camera typically produces a terabyte of data each day. The company’s products are used in close-in photogrammetry as well, which can be used to make accurate 3-D measurements of such things as the configurations of rooms and factories where the actual layout may differ from the original plans. The aerial variety can be considered a special case of close-in photogrammetry, and the company is working to produce software that will handle both.

Vexcel will use its technology to supply the content needed for Microsoft Virtual Earth and will continue to sell products to third parties. When asked whether this includes companies involved in supplying information to Google Earth and other Microsoft competitors, Gail noted that there are no restrictions on the use of the products.

Google, Microsoft and the other companies involved are not revealing product plans in any detail. They are working on improving the resolution and quality of the images, as Google did with its latest release. Plans to add radar-derived information to the data collected optically have been announced.

The software must render the images fast enough to be useful. That requirement implies a certain processing power and places restrictions on the bandwidth of the connection. By acquiring data and transmitting only what has changed from one view to the next, some of the bandwidth demands can be lessened. The drawback is greater computational needs.

How such demands are balanced against each other will be decided by each software firm. As for why the software is being offered, Google representative Megan Quinn noted that the company’s mission is to organize the world’s information. Geospatial imagery provides users with more information about the Earth. “Providing even higher quality satellite imagery will further this discovery,” she said.

Profitability

Besides the thrill of discovery, however, there are other reasons to increase the quality. Some serious money could be made by becoming the tool of choice for providing images. Geospatial images are considered crucial to the local search market, which is expected to be tens of billions of dollars annually. Some of the money will flow to Google, Microsoft, Yahoo! or Mapquest, but not all of it will.

One goal of all of these imagery viewing efforts is the development of platforms or venues, said Allen Weiner, an analyst with technology consultant company Gartner Inc. in Stamford, Conn. These platforms become something on which to build other applications.

A hotel chain, for example, might show shots of the area near its beachfront locations as a way to supplement standard map and driving directions. Potential guests could then see what is nearby and what the neighborhood looks like. That could help attract business to the hotel chain.

That may be commercially worthwhile, but Weiner noted that providing such basic information doesn’t require anything as futuristic as a live image. It may not even need any higher resolution than is currently possible. “I’m not sure what the huge gain is there,” he said.

On the other hand, GeoEye’s Brender thinks that better resolution will be commercially successful, and predicts that the optimum resolution will arrive with the next generation of satellites. “We think that at half-meter resolution, you’re hitting the sweet spot of what people will want,” he said.