Version 17 (modified by wja, 3 weeks ago) (diff)

--

Generating Photogrammetry Products

This page aims to guide a user through photogrammetry processing using Agisoft Metashape (formally PhotoScan).
Refer to ~arsf/doc/photoscan_processing/ for a step by step guide using a Metashape GUI.

Processing using Metashape

Overview: The following files comprise the arf_photoscan library:

arf_photoscan/
init.py
ps_export.py
ps_processing.py
ps_tools.py
ps_tools.pyc
ps_ortho_processing_after_input_gcp.py
ps_ortho_processing_before_input_gcp.py
README.md
templates/
ps_workflow.cfg

The arf_photoscan library currently contains two scripts for automating Metashape processing of NERC-ARF data: ps_ortho_processing_before_input_gcp.py will set up a Metashape project and align photographs. If Ground Control Points (GCPs) are available, then the script will stop and prompt you to manually locate them in the Metashape project. If no GCPs are available, the script will complete the whole Metashape processing workflow without (obviously accuracy of data is significantly reduced). This step is not automated. ps_ortho_processing_after_input_gcp.py is run after the and placement of Ground Control Points.

Ground Control Points
Coordinates of GCPs in CSV format can be added to Metashape. The user will then need to 'drag and drop' the marker locations to their locations visible in the photographs input. This needs to be done for all photographs each GCP is visible in.
Note: once a couple of GCPs have been located in multiple respective photographs; the 'reproject' button (looks like a recycling symbol) will automatically place the remaining photo. These will still need to be accurately places, but will save a lot of time nevertheless.

Running a script:

  • Update config file (template is located in /templates/ps_workflow.cfg).
  • As airborne, in the rsg-photoscan virtual machine, run:
/opt/photoscan/photoscan.sh -r ps_ortho_processing_before_input_gcp.py -c [CONFIG]
  • If no GCPs have been input in config file, (gcp_csv_path = None) or if gcp_csv_path does not exist, then ps_ortho_processing_before_input_gcp.py will run through the whole Metashape workflow without GCPs.
  • If a path to GCP CSV has been input into the config file, these GCP coordinates will need to be manually positioned to their correct locations on the aligned photographs. Once this has been achieved, the second script can be run:
/opt/photoscan/photoscan.sh -r ps_ortho_processing_before_after_gcp.py -c [CONFIG]


Config Information
The same config template is used.
The project directory (defined in config file) is the location of the Metashape project file (*.psx) and other data produced whilst processing.
The output directory (also defined in config) is where the Point Cloud, DEM and/or Orthomosaic is exported, if the associated parameters are True in config file. If the output directory does not exist, then it will be created.

Quality Checks
To assess the quality of the exported data, compare the orthomosaic to Fenix/Owl data, or any higher spatial resolution imagery that you have access to. Use this to assess if the X and Y accuracy is suitable.
For the Z (vertical) accuracy, you should compare the generated DEM to any points of known elevation at the time of flight but not the GCPs used in processing. For UK projects, there is good coverage of LiDAR data which may be able to make a good comparison. You can calculate the bias and RMSE between the processed elevations and the expected elevations (inferred from proxy data). Any statistical metrics calculated should be reported in the delivery's Read_Me.
If the accuracy of the data is questionable, investigate whether this is caused by the following:

  • GCPs
    • GCPs need to be frequent, well spread, precise and in the correct units. Also consider reprojection errors if the GCPs and/or exported data are in contrasting projections.
  • Eventfile
    • Check the values in the eventfile. There may have been a problem when extracted them from the navigation data.
  • Processing
    • Check that the GCP placement in Metashape has been undertaken in as many photographs that GCPs are present in as possible. Also look at the residual error of the project's spatial fit. Unusually high error values for a particular GCP marker would suggest that it has been placed in the wrong location.

Creating Delivery

This additional processing service will be delivered as additional items in the otherwise conventional digital camera delivery using the following structure:

Dirs Sub dir Example Files Comments
ES17_126-2017_170.kml Google earth file containing Fenix/Owl/camera positions
Read_Me-20171018.pdf Instructions for use and description of dataset, plus any problems encountered
Read_Me-20180222_orthomosaic.pdf Instructions for use and description of photogrammetry products
dem Digital Surface Model(s) produced from Metashape.
PHASEONE-IXU-ES17_126-2017170_dsm-utm30n.dem Digital Surface Model in ENVI format
PHASEONE-IXU-ES17_126-2017170_dsm-utm30n.hdr Header for accompanying Digital Surface Model in ENVI format
doc Useful documents
camera_data_quality_report2012.pdf Camera data quality report
ES17_126-2017170_ground_gps_points_composite.csv Ground control point (supplied by PI) used during Metashape Processing
PHASEONE-IXU-ES17_126-2017170_photoscan_processing_report_20180222.pdf Agisoft Metashape generated processing report.
RCD105_SN21_Calibration_Certificate.pdf Camera Calibration Certificate
eventfile PHASEONE-IXU-ES17_126-2017170-ImageEvents.csv CSV file containing pos/att info per photograph event (may be missing if camera crashed)
orthomosaic PHASEONE-IXU-ES17_126-2017170_mosaic-utm30n.tif Orthorectified mosaic of photographs in GTiff format
photographs photographs/PHASEONE-IXU-ES17_126-2017170-00001.tif Tagged tif files of each photograph in area of interest
point_cloud Exports of the 'Dense Cloud' generated in Metashape
ascii PHASEONE-IXU-ES17_126-2017170_point_cloud.txt Point cloud in ASCII format
las1.2 PHASEONE-IXU-ES17_126-2017170_point_cloud.laz Point cloud in LAS 1.2 format
thumbnails photographs/PHASEONE-IXU-ES17_126-2017170-00001.jpg JPEG thumbnails of each photograph in area of interest