We do a significant amount of Research and Development (R&D) in drone mapping. Our AV-900 rotary wing system has an L1/L2 Global Navigation Satellite System (GNSS) that we run in the Post-Processed Kinematic mode. This gives us the location of the AV-900 Survey System antenna to within a few centimeters relative to a local base station. This technique is commonly called Direct Geopositioning since, in theory, you can generate accurate mapping products with minimal ground control points.
It is very important to realize that direct geopositioning is not at all the same as having Ground Control Points (GCP). In direct geopositioning, we are estimating the focal center of the camera using offsets from the GNSS antenna (so called “lever arms”). Thus we know the approximate location of the camera at the time of the exposure (this also requires a trigger signal called the Mid Exposure Pulse, MEP, which is a whole other story!). We do not know the orientation (pitch, yaw and roll) unless we have augmented our GNSS with an orientation measuring device. With GCPs, we know the correspondence between image points and ground points from which we derive the location andorientation of the camera at the time of exposures. I use a rule of thumb that if all goes right, your lever arms are exact and you are using a camera gimbal, direct geopositioning will be within about a factor of 3 of the accuracy of using well laid out ground control (that is, the GCP project will have approximately 3x the accuracy of the PPK project). Now these are really just rules of thumb so do not take them literally. The most accurate configuration is direct geopositioning augmented by GCPs since this gives not only a close a priori estimate of the camera location at each exposure (from the GNSS PPK) but also ground points to tie down the model.
We are assembling some best practices for small Unmanned Aerial Systems (sUAS) or drone mapping.