Imagery processing:

intro to the assignment

GIS/MEA 584:

Mapping and Analysis Using UAS

Justyna Jeziorska

Center for Geospatial Analytics
North Carolina State University

Possible issues

Processing data from UAS can be challenging.
Sources for problems in the processing include:

• Data specific issues caused by:
• bad weather,
• failure of hardware (UAS or sensor),
• wrongly chosen sensor or wrong settings
• not enough information about image localization
• challenging terrain,
• bad flight planning

Possible issues

• Software specific issues caused by:
• incopability between UAS and software (like file extensions)
• memory problems or lack of computational power
• limitation of the software algorithms

Possible issues: bad weather

• Strong wing causes movement of the UAS (especially fixed wing) and the imagery can be very oblique
• Fog, drizzle or snow can limit the visibility and result in low quality images
• Direct sunshine can produce cast shadows that will interfere with processing

Solution

• None. Repeat the survey in better weather

Possible issues: failure of hardware

• UAS can crash
• Camera can not be taking photos (due to battery filure, interrupted connection with UAS...)
• IMU can stop logging the pictures

Solution

• Usually - none. Repeat the survey.
• Unless you are extremely lucky and UAS crashed after acquiring data you needed and you will be able to retrieve it.
• Or the camera stopped working just for a while and its not so significant to your data because you still have sufficient overlap without missing pictures
• If only log is missing - see next slide

Possible issues: not enough information about image localization

• It is possible to process data based on the geotags provided by flight log or EXIF data, but:
• geotag contains image location, not the location of the object depicted on photo
• It is therefore not possible to properly georeference the model without additional information
• without any information about imagery localization it is not possible to georeference or scale the outputs
  • (you can produce not georeferenced 3D model or use scale bars for scaling the model)

Possible issues: not enough information about image localization

Solution

• Use GCPs Place targets and include them in processing - more info how to process data with GCPs in Agisoft Photoscan in next slides
• Use RTK/PPK equipped UAS - more info here

Optimize Camera Alignment

Possible issues: challenging terrain

• Homogenous surfaces cannot be processed by SfM algorithms
• Large water areas cannot be processed either (still water is too homogenous and waves move)
• Moving objects can be depicted on multiple photos

Solution

• Avoid large water areas or mask them
• Avoid large moving objects or mask them
• (great tutorial on how to work with masks in Agisoft Photoscan)
• Aviod homogenous surfaces or set GCPs (or other objects) on them

Possible issues: lack of memory

Agisoft Photoscan memory requirements

Aligning Photos

Building Model (Height-field mode)

Possible issues: lack of memory

Agisoft Photoscan memory requirements

Building Model (Arbitrary mode)

Decimating Model

Possible issues: log file

• if only .jxl is available use script by Vaclav Petras

Possible issues:

... are endless