Ground projections of photo's used in reconstruction

Hi guys,

Thanks for any comments or help in advance for my weird problem. I have a set of images processed with ODM for my research project. It works great and all products (dsm,dtm,ortho, ect) look awesome, thanks for this great product to everyone who made this possible.

What I am trying to accomplish is to get a representation in Qgis of what has been captured by each click of the camera. The exif data on the raw images is not acurate enough for my purposes. In order to get a more acurate representation I created a python script that uses the reconstruction.json in the opensfm directory. This script reads the translation and rotation of each shot. From that I calculate the projection from the camera and where the corner rays hit the ground. (for testing purposes I have used a perfectly flat terrain). Then I translate the xyz coordinates to GPS coordinates create a shape file and load it in Qgis.

However… the projection I see in Qgis is slightly off with what is actually in the picture ± 5-10% and not consistent (see picture). In yellow what I calculated the projection to be. The photo is where it should be’ish’ if I align it with the orthophoto.

My question is: Probably I’m doing something wrong but just to be sure, this reconstruction.json is what contains the estimations of the camera positions right? :thinking:

Any other solutions to accomplish the same end results would be welcome aswell ofcoure :nerd_face:


Hey @Josh, could you share your code somewhere? It might be easier to test and provide suggestions that way.

1 Like

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.