Jeffrey LewisWonk School: Overhead Images

Peter Zimmerman, late of King’s College London, sends along a challenge that I am calling “wonk school.”

Peter proposes a series of take-home exams to see how well various readers do analyzing overhead images. And, as an inducement, he is offering to buy a pint of (good) beer for the best answer to the exam at a tavern of my choosing in Dupont Circle.

This is a take-home examination; There is an honor code. Send answers to: peter.zimmerman [at] cox.net.

The full text of the assignment is in the comment section.

Comments

  1. Andrew Foland (History)

    Can I suggest a twist for a followup assignment? A photoshop detection challenge, in which one image is extensively photoshopped and two others not, then all three are downsampled, transcoded into a video file, played on full screen size, screen cap’ed, and then saved via MS Paint (lossy) JPEG? (This likely approximates the processing path of the presentation still images we’ve seen.)

    I suspect this exercise would help people form an appropriate level of confidence in claims that various of the images have been “obviously photoshopped”.

  2. Jeffrey Lewis (History)

    Here is the full text of Peter’s assignment

    I have been watching with great interest as so many Wonkers attempt to do analysis of satellite imagery using Google Earth and similar sources. Not to complain about Google Earth; the pictures are often very good, but sometimes they are perplexing, as when – to get a pretty picture – the computer matrixes together several pictures without making clear when one moves from a picture taken on one date to one taken earlier or later. It was particularly fun with the old shot of central London where the shadow of Nelson’s Column pointed one direction, while that from the “Big Ben” tower of the Houses of Parliament pointed in a direction about 60 degrees away. And it was only with great care that I could find the boundary along Whitehall.

    That said, I am going to begin posting a series of Google Earth images of various interesting places from around the world, all of spots of national security/arms control interest, but many sites more of historic interest rather than current. In general, they’ll be of places I’ve been and where I have ground truth. My challenge to you will be to analyze the pictures both quantitatively and qualitatively. Unfortunately, some sites have been annotated by the Google community, so in order to play the game fairly you’ll have to turn off all levels. That includes the “roads” overlay. Also, no fair Googling Peenemuende or any of the museums, etc that are present today. You have to work from the image alone, and from references such as Jane’s Fighting Ships (or the Wikipedia equivalent).

    The first picture is of Peenemuende, Germany, site of the World War II V-1 (Fieseler 103) and V-2 (A-4) test site. For this exercise you may use collateral information so long as you don’t turn to maps of the site, either found on the Web or in books. You have to work from the image alone, just as the analysts working for R.V. Jones did.

    I note that this picture is far clearer than what was available in World War II and has the added advantage of color. Analysts have to work fast because the volume of imagery is so great. You should be able to answer my questions in one hour of work, and must not take any more than two hours.

    Type “Peenemunde, Germany” into the search box of Google Earth. Adjust your “Eye Altitude” to about 14 kilometers. Keep this ‘full’ picture in mind as I will refer to it in one of my questions.

    Peenemuende is the tip of the island to the right of center. Center your picture on the SE tip of the main runway in view. Adjust Eye Alt to 7 kilometers.

    Within the picture you now have in front of you, you should be able to locate:

    • One A-4 (V-2) rocket
    • One Fi-103 (V-1) cruise missile
    • The original launch track for the Fi-103 as used in WW2
    • Pruefstand VII (Test Stand VII), the launch point for A-4 rockets
    • Two East German naval vessels
    • A variety of East German aircraft

    Give the latitude and longitude of each of the above, although you don’t have to find all the aircraft. But do identify a few.

    Identify the type and class of each of the warships.

    Estimate the angle of the sun above the horizon.

    Estimate the time of day at which the picture was taken.

    Identify areas of concrete that date from WW2 (just a few of the many)

    Identify any railroads in the picture; estimate their age and whether they were in use at the time the image was acquired.

    Can you find any buildings that you can probably associate with the World War II base? Give Lat and Lon for all of the above that you find, of course. Please use degrees, minutes and seconds.

    The airfield was used by the DDR Luftwaffe. Tell me what the main purpose of the base was (fighter, recon, bomber, etc.); justify your answer; estimate the number of major types of aircraft stationed (eg 25 fighters, 13 helicopters or whatever)(don’t bother to try to identify the specific types (eg MiG-19, etc). That’s because the last major type is no longer present anywhere in the photo.

    What models of aircraft can you identify unambiguously?

    What was the launch direction of the Fi-103 track?

    Where is the nature preserve?

    Now, back out to the 14km eye altitude version. Within this picture are two electric power generating stations. Find them. Tell me
    what kind of fuels are used in each. Tell me crudely when each was built. Tell me if either is still in operation.

    “Show all work.” You can post comments and such on Wonk, but don’t spoil it for everybody by posting solutions. Send those directly to me at peter.zimmerman [at] cox.net. One good pint of beer at a tavern near Jeffrey’s office to the best overall solution. In the event of close calls, Jeffrey will assist me in judging.

    In about a week I’ll open the sealed envelope and give you the answers. Of course, I did all the analysis myself just as I’ve asked you to do it last week. My knowledge of the lay of the land from a visit some years back did me absolutely no good until I already had the answers in hand.

    If Jeffrey permits, in about two weeks I’ll post another image and make similar assignments. The one I’m thinking of will likely be a contemporary site.

  3. peter zimmerman (History)

    Andrew’s idea isn’t bad. I just don’t have the software to do it myself.

  4. Arnold (History)

    This sounds suspiciously like a work assignment one would give prospective research assistants (if open source photo analysis was part of the job description).

  5. SQ

    This ain’t wonk school; this is squint school. Besides, some of here us have day jobs…

  6. peter zimmerman (History)

    On second thought I will withdraw the challenge to find the V-1 missile in the picture. It is simply too hard because of its setting and the shadows, etc. The rest stands.

    -pz

  7. J House (History)

    Note the article today re the Chinese bases on Hainan-
    http://www.telegraph.co.uk/news/majornews/1917167/Chinese-build-secret-nuclear-submarine-base.html
    They didn’t point out 2 other tunnel entrances ~6 km north of those in the image (under 30 meters of rock), with additional piers under construction.

  8. peter zimmerman (History)

    It isn’t squint school. But it is a test of identification skills and pattern recognition. And it isn’t really a test for a research assistant, tho’ if I had the money it would be.

    Really, I’m trying to get across some fundamentals of military image analysis, because in the past I have seen far too many bungled analyses of everything from missile bases to nuclear reactors, and I thought some of you Wonks would like to play the game the way a professional would. Note: I notably bungled one that even made it on CBS evening news, to my shame. But I learned my lesson.

    I’ve gotten two submissions so far, one of which was pretty damn good and the other a good first stab.

  9. J.B. Zimmerman

    Argh, the Google Earth imagery of Hainan has cloud cover over the two northern piers. The 094 isn’t at the southern pier anymore (image just says ‘2008’). The ‘tunnel entrance’ site shows what looks like a large tunnel arch, but doesn’t look like said tunnel is flooded – if the tunnel is meant to contain subs, perhaps they haven’t opened the excavation to the sea yet? Or is the inlet hidden?

  10. Mark

    We all know how the video games of yesterday were really training future soldiers to fly drones, but little did I suspect that my “Where’s Waldo” books were training me to be a satellite imagery analyst.

    I found the V-2, but that’s all I have the patience for. I’m all about the low-hanging fruit.

  11. CosmicRay (History)

    I like armscontrolwonk because it deals with real issues. Id rather play Grand Thief Auto IV than do make-believe things here.

  12. Yossi, Jerusalem

    I think the wonk school idea is great and having PZ giving the first course on photo analysis is an excellent idea. Hopefully next courses will be on other interesting subjects for example:

    * history of IAEA and NPT

    * nuclear reactor construction

    * missile theory and practice

    * history of the Iranian nuclear program

    * the evil side of the CIA

    PZ assignments sound difficult and even scary. I didn’t dare yet to look at the first one but this is not rational. We don’t have to do them and bad grades wouldn’t damage our careers. We could just wait for and study the official solutions and no doubt learn useful and interesting things from them. However, doing the assignments will probably make us better analysts.

    Being more professional arms control wonks may not improve our finances but is a chance to do something good for this planet. Arms control is now mainly a diplomatic weapon of the US but the times are changing. One day all nuclear weapons will be under IAEA supervision and their owners wouldn’t be able to detonate them without UN approval which will be given only in cases of last resort self defense. To bring this day we need a strong and very professional arms control community.

    I’m going now to look at this bloody WWII German missile base.

  13. Peter (History)

    Well, I spent a good amount of time poring over Peenemuende, and I didn’t get very far. I’ll be happy if I got anything correct, really 🙂

    Full disclosure – I’m just an interested bystander.

  14. Sean O'Connor (History)

    J.B., the problem you’re having is that particular image on Hainan that is in Google Earth was taken in 2006. The 2008 date you see on-screen is the copyright label. People see those dates on-screen and get confused, thinking that they’re looking at imagery that is far newer than it may actually be.

    As for the imagery exercise, I for one thought it was great fun!

  15. MH

    Second place should be a few free rounds of Big Buck Hunter @ BH

  16. peter Zimmerman (History)


    PZ assignments sound difficult and even scary.” said Yossi.

    The assignment isn’t difficult, really, except for the V-1, which is phenomenally hard to find, and since nothing is at stake, whatcha got to lose? Nothing. But what you might gain is an appreciation of the difficulty of actually finding something from above and the equal problem, once you find it, of identifying it. Oh, and a good beer is to be gained as well.

    Easy? The two ships are where ships have to be: in the water and tied up at a pier. Need any more hints?

  17. Andy (History)

    Well, if Sean doesn’t win this thing, I’ll be quite surprised.

    But this exercise is serendipitously similar to what I’ve been reading and researching lately, which is cognition and perception in relation to intelligence. Analysis of information – even imagery – is highly dependent upon context, cognitive bias and perception. An excellent overview of the topic can be found in Perception and Misperception in International Politics by Robert Jervis published back in 1974. Here’s an example that relates to the topic at hand:

    …British intelligence analysts were told that V-2 rockets were launched off a “steel cone surrounded by a square framework,” they examined pictures of a German test area and “found no fewer than 12 large conical objects, 15 feet in diameter,” which they took to be the platforms. When the objects later disappeared and the analysts calculated that they could not have been lifted over the surrounding wall or taken out through the narrow gap in it, they correctly realized the “conical objects” were standard tents, which they had often seen before, which had been folded up and removed.

    and

    In many cases, discrepant information is simply not noticed….And in World War II, a British photographic reconnaissance analyst who thought the German secret rockets would be a “70-ton monster, launched only from enormous rail-served projectors,” studied pictures of a German facility several hundred yards from the nearest railway line and paid little attention to what he saw as “a thick vertical column about 40 feet high and four feed thick.” The “column” was actually an erected V-2.

    Both are, of course, classic cases of cognitive bias affecting analysis.

    Sadly, I’m traveling now, so won’t participate in this current contest, but I will certainly be interested in the results.

  18. Martin Dirksen

    Dear Sirs,
    Thank You much for the assignment. As I wrote before in a comment: there is a definite need for “ Open source intelligence” , this also includes the usage and knowledge of the (cautious) interpretation of satellite-pictures . And to those critics out there: nobody has to take the exam.
    Yours ( from Hamburg, Germany, not to far away from Peenemünde,)
    Martin Dirksen

  19. J.B. Zimmerman

    Sean- Ah. Thanks. That makes sense then.

  20. Lao Tao Ren (History)

    How about photos from the Cuban Missile Crisis for one of the lessons?

  21. FOARP (History)

    Damn, wish I’d 1) seen this earlier, and 2) didn’t have an exam tomorrow – will there be a second installment?