Forum

Author Topic: difference in sparse and dense pointcloud extent  (Read 1166 times)

EV_91

  • Newbie
  • *
  • Posts: 4
    • View Profile
difference in sparse and dense pointcloud extent
« on: June 17, 2019, 01:43:01 PM »
Hi all,

I am working with RGB imagery for crop monitoring in agricultural fields. From time to time when I process the images into an orthomosaic there is a problem with the extent of the dense cloud and the subsequent data products (DEM and Orthomosaic). See the attached image for an example of the problem. The sparse cloud extent matches the extent of the imagery, while in the dense cloud a part of the scene is missing. So far I haven't found the cause of this problem and a way to solve the issue.

Do you have any toughts on how to resolve this?

Thanks in advance

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 13289
    • View Profile
Re: difference in sparse and dense pointcloud extent
« Reply #1 on: June 17, 2019, 03:44:36 PM »
Hello EV_91,

Can you please check, if the bounding box (Region) does include all the area of interest? If not, it should be manually adjusted prior to the dense cloud generation.

Also missing dense cloud areas could be caused by the lack of the image overlap (low number of matching points for the image pairs), but such areas can be interpolated - if you use Extrapolated option in the Build DEM dialog, the surface would be extrapolated up to the bounding box sides, therefore you will be able to generate the orthomosaic for the whole area.
Best regards,
Alexey Pasumansky,
Agisoft LLC

EV_91

  • Newbie
  • *
  • Posts: 4
    • View Profile
Re: difference in sparse and dense pointcloud extent
« Reply #2 on: June 17, 2019, 04:27:56 PM »
Thanks for the fast response Alexey,

indeed the region is too small, this caused the incomplete dense cloud generation. Is there a way of importing a shape (plot boundaries are stored in shapefiles) and setting it as region object? Normally I process my imagery programmatically using the python API so this would be much more convenient than adjusting it by hand.


Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 13289
    • View Profile
Re: difference in sparse and dense pointcloud extent
« Reply #3 on: June 20, 2019, 08:02:48 PM »
Hello EV_91,

The bounding box can be defined via Python script - chunk.region instance can be defined via region.center, region.size and region.rotation. Note that all these attributes are defined in the internal chunk coordinate system, so if you are processing georeferenced data you need to perform additional transformation to get proper values in the internal coordinates for the region definition.

There are some threads with the code examples that are resizing/relocating and re-orienting the bounding box for georeferenced chunks:
https://www.agisoft.com/forum/index.php?topic=10401
https://www.agisoft.com/forum/index.php?topic=7543
https://www.agisoft.com/forum/index.php?topic=8679
If you need some assistance for your project, let me know. The easiest approach, I think, would be just increasing the region.size by certain factor, for example: chunk.region.size = 1.25 * chunk.region.size
this should increase all the sides by 25%.
Best regards,
Alexey Pasumansky,
Agisoft LLC