Cylindrical pans to polar stereographic projections?

tim53tim53 Global Mapper UserPosts: 68Trusted User
edited November 2012 in Projection Questions
Hi again...

I'm interested in taking 360 degree panoramas from the Opportunity rover on Mars and reprojecting them as polar stereographic overhead projections, that I would then georeference to the 25cm/pixel orbiter image base map of the landing site.

Currently, I'm using Photoshop's "polar distortion filter" to make simple polar projections, but these are compressed toward the horizon and stretched near the rover. I'd prefer a polar stereographic projection out to a certain radius from the center that preserves the shape of objects - e.g., small craters would be circles. Currently, it can take 2-400 tiepoints to georeference the simple polar projections onto the orbiter orthographic image. I may be wrong, but it would seem that it ought to be easier and quicker if I could start with polar stereographic projections.

I don't see anything like a cylindrical projection, where the x-axis is azimuth and the y is elevation, in Global Mapper. So I'm assuming I need to treat the cylindrical pans as if they were equirectangular maps and create tiepoints where lat/lon is elevation/azimuth from the original file. So far, I've gotten wildly distorted results, but maybe I just need to input a bunch of tiepoints, rather than 4 or 6?

-Tim.

Comments

  • global_mapperglobal_mapper Administrator Posts: 17,238
    edited November 2012
    Tim,

    I can't imagine the results of this would be particularly good since the images from the rover would only have pixels for ground directly visible from the rover and there isn't necessarily a direct correlation of location in the image to distance from the rover, particularly near the horizon where a distance mountain might be just a few pixels over the surface only a few kilometers away. You would definitely need more tie points, but even with one for each pixel you would get weird stretching and warping due to issues like this. I can imagine an absolute worst case if you were somewhere with say an arch and the associated tie points below and above the arch were some far distance then the arch itself was close. The image would get warped all over itself trying to map that to a top-down view.

    Thanks,

    Mike
    Global Mapper Guru
    gmsupport@bluemarblegeo.com
    http://www.globalmapper.com
  • tim53tim53 Global Mapper User Posts: 68Trusted User
    edited November 2012
    ...that was when I heard the term "2.5D" projection for the first time. Not a 3D view, because that could involve situations with multiple z values for each x,y pair.

    I've been experimenting with tying polar projections, first to the orthographic projections (that are projections of the cylindrical pan onto a DTM generated from the stereo from that pan), which extend 15 meters from the rover, then to the orthographic orbiter image. It's surprising how well it works, but it only works where the terrain is visible from the rover (of course) and it takes a long time to add all the tie points. I don't try to tie terrain beyond ridges. Fortunately, the topography at the Opportunity site is almost never so complex that I'd have arches to deal with. It's pretty flat, at least out to a few tens of meters, typically. I'm also working with our software folks here to extend the DTM beyond 15 meters radius, based on the stereo data, perhaps by "allowing" lower quality correlations at greater distances. An obvious solution would be to take stereo panoramas more frequently so that the "standard" orthographic projections overlap. But there's always a trade-off between covering more ground and getting data down.

    On Earth, one can always rent a 4x4, grab their camera and tripod gear, and go out to a field site and fill in any gaps that might exist. But I can't do that on Mars! ;)

    -Tim.
  • global_mapperglobal_mapper Administrator Posts: 17,238
    edited November 2012
    I'm guessing that the imagery from the satellites around Mars isn't nearly good enough for what you are after? Man it would be great if the Observer had a Lidar instrument with color so we could easily get a true 3D representation of the surrounding terrain so a real 3D model could easily be built. Then we would have a location for each sample/pixel already baked into the data.

    Thanks,

    Mike
    Global Mapper Guru
    gmsupport@bluemarblegeo.com
    http://www.globalmapper.com
  • tim53tim53 Global Mapper User Posts: 68Trusted User
    edited November 2012
    I'm guessing that the imagery from the satellites around Mars isn't nearly good enough for what you are after?

    The highest resolution images we have from orbit are from HiRISE, map projected to 25cm/pixel. Since we needed all the images we could get to select the landing site, we also have good stereo. USGS Flagstaff made DTMs from this stereo with 1-meter postings. The navigation camera pans from the rover are map-projected at about 1cm/pixel. It'd be nice to have data in between, or to be able to extend the navcam DTM farther from the rover (which is possible if the viewing angle is optimal) without moving the rover to do it (though that's also a possibility, so long as we're not driving long distances in a sol).
    Man it would be great if the Observer had a Lidar instrument with color so we could easily get a true 3D representation of the surrounding terrain so a real 3D model could easily be built. Then we would have a location for each sample/pixel already baked into the data.

    The Mars Laser Altimeter had a footprint of about 150 meters on the ground, and shots were ~300 meters apart on center. Ground tracks were up to a couple of kilometers apart at the equator.

    -Tim.
Sign In or Register to comment.