In structure from motion process, the software is able to generate a final report where it outputs the average value of the GSD, the Ground Sampling Distance.
It’s calculated based on of all the images that you shot on the field and and make the dataset up.
GSD is a quite important factor that depends on:
- camera sensor (size and resolution);
- focal length (real focal length, not the 35mm equivalent one);
- distance of the camera from what you have photographed.
If you run a structure from motion process using pictures taken at a great distance you’ll have an average GSD that will be greater than the one calculated from just closer images.
If you mix up images shot at different distances you’ll have back an average GSD.
It’s gonna be higher than local GSD based on some pictures (the closest ones) but it’s gonna be lower that local value calculated on the others (the farest ones).
Try not to get stuck on the average GSD but don’t skip it either.
It is a good indicator of your aquired data.
The SfM software works to build the point cloud using all the images, the detailed and the less detailed ones, but then it does a mathematical average of all the GSD values.
You have differences of GSD even for each pictures of the dataset.
Espeacially if you have images of vertical objects.
I think that the most relevant data, in this back analysis, is the general accuracy calculated on the information of the check points.