mosaicing old aerial photos

[maranhao6456]

Hi! My name is Marcelo Maranhão and I work at IBGE - Brazilian Institute of Geography and Statistics, where we are developing a project to catalog and digitize old aerial photos. I would like to know if there is anyone here in the group working with old aerial photos. I’m testing WebODM to mosaic a large set of photos.

[Gordon]

Welcome to the community Marcelo

You might be interested in this thread, and the 2018 one linked within:

Processing historical aerial photos, revisited

Hello, I’m revisiting a project similar to one I discussed in this 2018 post, where I had successfully generated a decent point cloud using low-precision Ground Control Points (GCPs). Back then, ODM was using PMVS instead of the powerful SMVS algorithm. I’m now working on a similar process with aerial photos from the same flight (60% forward overlap, 20-30% side overlap), but covering a different area. I’m using GNSS data generated from double and triple frequency DIY receivers, with millimetr…

[geofis]

Hello Marcelo and Gordon,

I’ve been using WebODM extensively for processing historical aerial photographs and would like to share my recent experiences:

1. In the thread, you will see that I recommend how to use ODM to achieve better results with historical aerial photographs. It is important to ensure the reconstruction is done well. Even if the reconstruction doesn’t work with GCPs, I have tried generating an orthomosaic with image coordinates and then georeferencing the orthomosaic using any georeferencer (for example, QGIS/GDAL).
2. WebODM is a powerful tool for photogrammetric processing of historical aerial photographs, but it’s essential to be aware of its limitations.
3. For generating orthomosaics, consumer-grade scanners are generally sufficient. You don’t need highly specialized equipment for this task.
4. However, if your goal is to create DSMs/DTMs, it is crucial to use the highest resolution scans possible (e.g., >1300 DPI) and professional-grade photogrammetry scanners to ensure accuracy.
5. When producing DSMs/DTMs, it’s beneficial to use additional software for camera calibration or improve calibration by directly editing the reconstruction. This could involve excluding points with high reprojection error, for which a point cloud viewer or software that applies exclusion criteria with thresholds would be useful.
6. If the scans are significantly distorted and you only need orthomosaics, pre-warping the images can be very helpful. This can be done using polynomial methods (preferable) or thin plate splines (if there is substantial original deformation in the scans). While using TPS can complicate obtaining accurate camera intrinsics, it tends to produce better orthomosaics. However, it may not help and could even worsen the DSM/DTM results.
7. I also recommend removing the background frame from the images, as it can sometimes prevent ODM from generating a suitable reconstruction.

I hope this information is helpful, and I’d be happy to discuss further if you have any questions or need more detailed insights.

José Martínez

[maranhao6456]

Hello Gordon and José,

Thanks for your answers!

José: Regarding item 1, I was wondering if it is possible to generate a mosaic without georeferencing the images and without GCPs. From what I researched on the web, in Metashape it is possible to generate mosaics in a preliminary way and without geometry.
In my attempt in ODM, using the default configuration, I was unable to generate the mosaic without the GCPs.
Is there any configuration in ODM that allows generating without the GCPs and without georeferencing the photos?

Regarding items 3 and 4, I am scanning using cameras on a light table, and working with slides and not paper photos. We saw a project in Portugal that works with this type of acquisition and had good results.

Marcelo Maranhão,