Catch up on stories from the past week (and beyond) at the Slashdot story archive

 



Forgot your password?
typodupeerror
×
Google Businesses The Internet Science Technology

How Google Earth Images Are Made 122

An anonymous reader writes "The Google Librarian Central site has up a piece by Mark Aubin, a Software Engineer who works on Google Earth. Aubin explains some of the process behind capturing satellite imagery for use with the product. 'Most people are surprised to learn that we have more than one source for our imagery. We collect it via airplane and satellite, but also just about any way you can imagine getting a camera above the Earth's surface: hot air balloons, model airplanes - even kites. The traditional aerial survey involves mounting a special gyroscopic, stabilized camera in the belly of an airplane and flying it at an elevation of between 15,000 feet and 30,000 feet, depending on the resolution of imagery you're interested in. As the plane takes a predefined route over the desired area, it forms a series of parallel lines with about 40 percent overlap between lines and 60 percent overlap in the direction of flight. This overlap of images is what provides us with enough detail to remove distortions caused by the varying shape of the Earth's surface.'
This discussion has been archived. No new comments can be posted.

How Google Earth Images Are Made

Comments Filter:
  • by creativeHavoc ( 1052138 ) on Sunday April 29, 2007 @10:52PM (#18923415) Homepage
    don't confuse google maps with google earth. They have distinct purposes, and excells at them.
  • by dickeya ( 733264 ) on Sunday April 29, 2007 @11:07PM (#18923533)
    Maybe you should download and try Google Earth which....

    requires 3D hardware, can do smooth zooms, tilts, and pans, and showa the world with elevation.

    Google Earth IS the evolved version of the Keyhole client you referred to.
  • by Anonymous Coward on Sunday April 29, 2007 @11:30PM (#18923671)
    I was one of the Aerometric-Alaska flight operators that took photography in 2006 while on-board a variety of small planes. The film we used was generally Kodak 2444, with 9' x 9' shots. After development, these prints can then be scanned at a resolution comparable to roughly that of an 11 megapixel camera. As the article states, these photos are usually taken in succession with 60% overlap. This is what has allowed people to generate topographic maps for decades, even before complex computer interpolation and computer graphics capabilities were present. Stereoscopic perspective of the same area of land taken from 2 separate angles allows people to determine differences in height, in case anyone has ever wondered how that worked. Nowdays, surveys and digital radar scanning is where most of the information that modern topography uses tends to come from.
  • by kefler ( 938387 ) on Sunday April 29, 2007 @11:32PM (#18923679)
    Actually there's no need for a camera's "panoramic" mode any more. Check out Autostitch [cs.ubc.ca], a free for personal use program created by researchers at UBC. Essentially you take as many pictures as you want with varying amount of overlap. Each picture can be rotated differently and even vary somewhat in exposure, and this program automatically figures out which ones go where, even throwing out ones that are not part of the scene. It takes a ton of ram and some CPU speed but the result is better than any other method I've seen. Some examples here at the bottom of this page: AZ Snow Pictures. [sabinovalley.com]
  • by dickeya ( 733264 ) on Sunday April 29, 2007 @11:46PM (#18923751)
    Unfortunately, when working with that kind of imagery collection the high tech part is on the processing end.

    You need to:
    1. Correct for lens distortion
    2. Correct for tilt
    3. Correct for terrain distortion
    4. Correct lighting imbalances across the scene
    5. Assign it bounding coordinates of a known mapping coordinate system / projection

    This is the basic process for making an orthophoto [wikipedia.org]. These are generally dealt with using a software package like Erdas Imagine which can deal with all the steps in one swoop. It looks at the lens info, coordinate tie down points, an elevation model and outputs a photo that can be used for linear measurements.

    So anyway, it is possible to accurately georeference many sources of imagery, it just depends how much time you want to spend processing it. If you plan on covering a large area, taking photos out of the window of a Cessna is probably not the best way.
  • by xenn ( 148389 ) on Monday April 30, 2007 @12:47AM (#18924021)

    Google put it up and at the end of the day, Sydney wasn't Sydney any more. Instead, Sydney was transformed into one big banner ad:
    that's funny, in the link you provided they say this:

    So, after hours of combing through the new images, we've found no real examples of private advertising or even any sign that the people of Sydney knew they'd be on Google Maps!
  • This is not new (Score:4, Informative)

    by ouzel ( 655571 ) on Monday April 30, 2007 @12:56AM (#18924055)
    The process he described is the same process that imaging companies have used since LONG before Google Earth acquired Keyhole. And many of those companies are still doing it. In fact, many of them are the same companies from which Google is now acquiring the imagery used in Google Earth. Does this guy really believe Google is conducting their own overflights and sending up their own balloons? Does Google now have their own satellites, too?
  • by BillGatesLoveChild ( 1046184 ) on Monday April 30, 2007 @01:18AM (#18924195) Journal
    Maybe Google's Army of ex-CIA Photointerpreters were working to a deadline? Or maybe she just means "we looked at a few". Seriously it did happen: Here are photo blogs of expectant advertisers (and one poor sucker who blew $10,000 on a sign they didn't fly over).

    http://blogs.smh.com.au/mashup/archives/009502.htm l [smh.com.au]
    http://www.smh.com.au/articles/2007/01/29/11699192 56978.html [smh.com.au]

    Though good luck to the guy from SOS Print+Media! Hahahahahha!

    Google say they're thinking about doing the same over parts of the US on Independence Day. Don't think they'll try it over the UK though: Too many swear words and phallasus. http://www.smh.com.au/news/web/google-spots--crap- circles/2007/02/01/1169919445548.html [smh.com.au]
  • by zippthorne ( 748122 ) on Monday April 30, 2007 @01:37AM (#18924271) Journal
    The camera's mode, at least in Canon's case, is to show an overlay of the pictures already taken in the matrix or panorama you're taking. It's easier to get good, overlapping, easy-to-stitch images if you use the camera's framework to help you set up the shots, and be sure when you've taken enough to fill the planned matrix. As a bonus, the pictures are all tagged as to their position in the final photograph, and all the camera data is recorded for the stitching program.

    IIRC, the actual stitching still happens in software on a PC.
  • Because (AFAIK) digital cameras don't exist in the large format form factor that [film] cameras for aerial photography uses. Even if they did, from typical aerial photography altitudes digital camera are inferior in resolution to existing films.
  • by hughk ( 248126 ) on Monday April 30, 2007 @04:44AM (#18925121) Journal

    Both require a GPS-controlled platform, capable of shooting several shots a second.
    Funny that, aerial mapping has been used for a looong time and before GPS. All you needed is a reference point with coordinates and then the rest follows. The old equipment used to put altitude, speed and direction onto the film for later use. Some cameras would take frames but some would in effect take a continuous strip using line-scan techniques.
  • by JayBat ( 617968 ) on Monday April 30, 2007 @05:41PM (#18933573)
    By the way, the thing that boggles me about all the mapping services out there is how they do routes: how do they determine where the roads are (DOT?), how do they store the roads, and how do they calculate driving routes (that often appear to take into account traffic speeds).

    A company called Navteq [navteq.com] does a lot of it, and contracts out data and software.

    The New Yorker had a great profile on E-mapping and route finding [newyorker.com] including a ride-along with a "Ground Truth" team that heads out with their GPS-linked laptop and drives... pretty much everywhere. One key part of ground-truthing (and good directions) is knowing the signage on the route:

    Singh bought a Red Bull and took the wheel. Arcari sat in back with the laptop, ready to note any changes in what they called the "geometry" of the roads.

    "Whenever you're ready, Shovie," he said.

    The first thing the men noticed was a "No Left Turn" sign out of the gas station. "That doesn't go in the database," Arcari said. "That's unofficial geometry, since it pertains to a private enterprise."

Top Ten Things Overheard At The ANSI C Draft Committee Meetings: (5) All right, who's the wiseguy who stuck this trigraph stuff in here?

Working...