Number of geotagged images in a set

Hi everybody

Is there a minimum required for number of geotagged images (or perhaps %) to make an accurate orthophoto or DTM ?


I have never had anything consistently work below ~16 for orthophoto and DTM. There are a lot of variables at play, and the fewer the images, the more the edge effects are going to creep into the core of the area of interest. 32 is what I would recommend consistently works, but watch those edges where there isn’t overlap and clip accordingly.

1 Like

Thank you. But the question was about what can happen if you have some geotagged images and other without geotag. Do ODM need a minimum % of geotagged images for good georeferenced result.

I guess it depend if geotagged images have a good geographical distribution but I don’t know if there is a minimum of geotags required.

Oh! Gotcha. I misunderstood the question. Hmm – that is a fantastic question. In theory, just a handful of images will work in principle, but keep in mind that much of the processing optimizations in OpenDroneMap (at this time) rely on geographic data, so you enter the territory of some very expensive structure from motion calculations without geographic tags. This could be fixed with more clever processing approaches that use bag-of-words or other feature matching approaches, but I don’t think we have any of these approaches exposed.

All that said (and without testing) a handful of images well distributed across the dataset should do the trick. But the details are important. It’d be interesting if you did some tests and report back. I don’t know if it’s been explored yet.

1 Like

I will test with different distribution and come back with the result