Agisoft Photoscan pro

This is a basic guide on how to work with the photos collected from your UAV flight. This guide is written for images collected with a Smartplanes SmartOne-C. More detailed guides can be found on Agisoft’s website.

Do the following settings before you start:

  • Open “Tools” –> “Preferences…”. In the OpenCL tab, deactivate one core for each graphics card you have installed. In the example below I have one graphics card:

agisoft_preferences_opencl

Brief overview of the workflow:

  1. Add your photos in Photoscan: “Workflow” –> “Add photos…”.
  2. Import the camera positions (if non-geotagged photos). The camera positions is imported by clicking “Import”-button in the “Reference”-window. Make sure that you choose the right columns for the files. If the file contains orientation data, check the “Load orientation” checkbox and choose the right columns. Choose the coordinate system that the camera positions are stored in.
  3. If your camera is rotated on the drone, you need to specify this under “Tools” –> “Camera Calibration…”. Under the tab “GPS/INS Offset”, specify “Yaw (deg)” to 90.
  4. Check the settings. Click the “Settings”-button in the “Reference”-window. If you’re using orientation data, we recommend that you increase the accuracy to 10-20 degrees if you’re not sure that you have much better accuracy than that.
  5. Align your photos: “Workflow” –> “Align photos…”. Set “Accuracy” to “High” and “Pair preselection” to “Reference” if your photos are geotagged or you have imported camera positions.
  6. Go through your photos and uncheck and disable photos in turns of the photo-block. After you have done this, run “Optimize alignment”.
  7. Use the manual selection tools to remove obvious outlier points in the sparse point-cloud. You can also use the “Gradual selection…”-tool under the “Edit”-menu.
  8. If your ground control points (GCPs) are in a different coordinate system to the images, you now need to transform your project to the same coordinate system as your GCPs before you import them. This is done by clicking the “Convert” button in the “Reference”-window.
  9. Import ground control points (GCPs) if you have. This is done by clicking the “Import”-button in the “Reference”-window. The following settings should be done if you import GNSS-coordinates from the Trimble GeoXR (and exported the coordinates in Sweref99 TM RH2000):
    importGPSPointsToPhotoScan
  10. Go through GCP by GCP and point them out in the photos. You can right-click on the GCP and choose “Filter photos by selection…”. You can switch between photo by using Page Up and Page Down. If you can’t place all GCPs, you can run the next step and then try again. The GCPs will be better placed after an optimization has been run.
  11. “Optimize alignment”. If your GCPs are of high accuracy, then you should uncheck all photos before you do an “Optimize alignment”. Make sure that all GCPs you want to use are checked.
  12. When the above steps are done and the error in the GCPs are small (~5-20 cm if using RTK GNSS receiver).
  13. Now you can either do a batch script doing the commands below or run them one by one.
  14. Build dense cloud.
  15. Export dense cloud as las for example.
  16. Build mesh (normally from the sparse cloud).
  17. Create ortophoto using the mesh from the sparse cloud.
  18. Export ortophoto as tif for example.