An example might be easier .. Say we have five nice reference points in the current model - intersection of tiles, edge of concrete, painted line on the road .. whatever - something you can pick in the photos .. 1,2,3,4,5. Let's say "3" is your known point and 1,2,4,5 are somewhere near the four corners. In the current model - known point "3" differs from the real world by 300mm South,175mm West and 20mm too low in elevation. By creating the tie points in the current model - all you are doing is getting the current coordinates for the features .. and when you export them you get a convenient text file to edit. Now edit the coordinates by 0.3,0.175,0.2 .. say .. so moving all the "reference" points the same amount. Now import those points back in as control and pick the SAME reference features in the images .. you are moving everything relatively - you haven't changed the AT at all - or if you have - only by how good your picks are - keeping in mind the GPS on the drone is +/- 50mm to start with .. the calibration from your GPS antenna to shutter is probably 50 or 100mm out .. then your IMU (for tilt) is probably good to half a degree if you are lucky .. and the compass is probably good to half a degree ..
↧