Automatic SFM processing

The Photo database features an automated process for generating SFM models using Agisoft Metashape. This process is customizable through the use of profiles, such as those for historical or thin section imagery.

Example


A quick (slightly awkward) walkthrough on how to setup up automatic metashape processing.

How do I launch automatic processing?

When the network scanner check the metadata.xlsx files during a scan, it will create Metashape projects if the string of characters startAutoProcess is present in the Processing row. Once a processing is launched, the string of characters will be changed to autoProcessStarted.

Example

autoProcess.png
An example of the contents of a metadata.xlsx file that will create metashape projects when scanned. Note the contents of the Processing row, as well as the autoProfile row - in this case the latter specifies a special profile, defining known camera properties.

A project will be created for every folder that contains jpg or tiff images. The project will be created in a folder called auto-project-# where # is the folder name that contains the images.
This applies to any subfolder with the exception of:

  • folders that have auto-project in their name,
  • folders that have .files in their name,
  • and folders that have a separate metadata.xlsx file in them.
    No projects will be generate for folders that have any of the above conditions met.
Example

auto-process-folder-structure.png
An example of the contents of a folder where automatic processing has started. Each of the folders starting with SN have tif images in them, therefore a folder was created containing the Metashape projects and associated outputs.

Auto process project profiles

A project profile can specified in the metadata.xslx file, by adding a row named autoProfile and writing one the profiles below in the Values column, see the first example on this page. Please note that case matters in all these!
If no profile is specified, a generic project will be created. Point clouds, DEMs and Orthomosaic features will only be created if the images are geo-referenced.
There are currently several profiles available:

  • thinSection: will create an Orthomosaic from a series of overlapping thin section images. A webpage containing he image will be created at the page using the name of the folder where the images were taken: https://alsdo.canterbury.ac.nz/SEE/autoOSD/ . This allows you to share the image over the web.
  • scanned: will create a generic scanned image project, attempting to detect fiducials. Point clouds, DEMs and Orthomosaic features will only be created if the images are geo-referenced. A series of scanned sub-categories can be specified, if you know the properties of the camera that took the historic imagery. You can find information about the camera properties is contained in the folder \\file\Shared\SEESPhotoDatabase\Archive\LINZ_Camera_calibration. Currently(06/2024) , profiles exist for these cameras:
    • Eagle354_210mm_1946
    • WILDRC5_115mm_1955
    • WILDRC8_210mm_1964
    • ZEISSRMK_150mm_1979
    • ZEISSRMK_305mm_1981

Summing it up

Key steps:
  • 1 folder of images in SEESPhotodatabase = 1 project
  • startAutoProcess anywhere in the Processing row of the metadata file
  • Use Agisoft network monitor project progress. (can also be accessed via http://seesalsdo1p:5840/ if you are on campus)
  • cameraCalibration and other profiles files are available - get in touch if you have a specific use case or need help!

Key information:
  • shared folder: \\file\Shared\SEESPhotodatabase\
  • server name: seesalsdo1p
  • metadata file name: metadata.xlsx