Problems stitching photos

Hello everyone,

We were asked to photograph maize crops for an R&D project, and we did it as we usually do for photogrammetry projects with a Canon SX230 shooting from an Ardupilot equipped wing at 75 meters.

We are having problem stitching the photos because all the software apps we have tested seem to get confused with such a uniform subject. A corn irrigation pivot has a radius of 300 meters, so we end up with more than 200 pictures of concentric green circles. So far we tried with Agisoft, Dronemapper and other non photogrammetric stitching applications without luck. Interestingly the best results we had so far  were with Photoshop.

Has anyone found similar problems? 

JL

You need to be a member of diydrones to add comments!

Join diydrones

Email me when people reply –

Replies

  • Hi, have you tried VisualSFM?

    I  did some flybys of potato trial crops the other day  to do some tests, though at only ~30m or so and using lower resolution images (1920x1080) from an old gopro, corrected using PTLens.

    VSFM worked quite nicely, as did MS ICE and AirphotoSE but I found that its all down to how much overlap you have with such a uniform crop (you need 50% overlap with each image at least I think).

    Good luck :)

    VisualSFM : A Visual Structure from Motion System
  • Have you tried using this feature of the mission planner?

    Geotagging Images with the Mission Planner

    Aaron

  • I don't know of a solution for your current image set, but a couple of things you could try
    if you fly the field again are:

    Place markers in several places through the field. We have used 2" white vinyl window blind strips for this. Place them in different configurations to form unique shapes. They need to be visible from the air of course.

    It may also help if you fly higher. Try 120m.

    Depending on your goals, the fact that you can't find significant variation between your images may actually be good information. It indicates a very high degree of uniformity, and the data from a few images could therefore be used to indicate conditions for the whole field.

    Deon
    course.it
    This domain may be for sale!
  • Hi JL,

     

    Just thinking a bit further, you could identify key features in the images, return to the site and measure those points, then add these as Ground Control Points, normally five points across the flight area, one in each corner and one in the middle is sufficient, but with your data you might want to do it every 50m around the edges and some in the centre?

     

    See the PhotoScan help file under "Place Markers" in Orthophoto & DEM Generation with Agisoft PhotoScan Professional.

     

    Best regards,

     

    Keith,

  • Hi JL,

    I assume that your system does not geotag the images?

    We have been working on crop imaging for two years now and the only way to prevent your current situation is to use a system which automatically geotags the images and records the pitch, yaw and roll. Some systems use Omega Phi and Kapa instead of pitch / yaw and roll.

    Another option is to use Ground Control Points, but this is normally used in conjunction with Geotagged images.

     

    Programs you could try with your current data are;

    EnsoMOSAIC by Mosaic Mill

    Rapid Station by PIEneering

    APS by Menci

    Pix4D

    or GeoID

     

    Some people have used ICE by Microsoft which is a free image stitching program;

    http://research.microsoft.com/en-us/um/redmond/groups/ivm/ice/

     

    Where are you located?

     

    I can assist you in selecting a good system for next year!

     

    Good luck,

     

    Keith,

     

This reply was deleted.