Poor Relative Accuracy of Orthomosaic without Ground Control

So I flew a high school football field to take photos of their new turf field. Airspace only allowed flight altitude of 150 AGL so I could not get the entire field in a true vertical shot due to those low flight restrictions so I decided to create an ortho to generate that vertical image. Everything came out great and I converted the GeoTiff to a JPEG. Then just for the heck of it I decided to bring the GeoTiff into QGIS and look at the relative accuracy of the ortho compared to the football field lines. This project was flown at 125 Ft AGL (0.4 in / px) with no ground control. To my amazement when I measured the 100 yard football field it only measured 93 yards in QGIS! Now I understand that with no ground control the spatial relevance of the project with be based the the autonomous GPS position plus or minus a few meters from the real world spatial coordinated. However, I though that the relative accuracy would be a lot tighter than 7 yards error in a 100 yard measurement. Am I missing something here?

Here is the GEOTiff:

I would put it in this sort of context. Drone GPS receiver typically would have a similar accuracy to a smartphone GPS given a good view of the sky.

Smartphone GPS could easily deliver 10m accuracy on the X and y coordinate (or north and east).

Therefore what you’ve measured is less than 10% error which I would say is pretty good.

Unless you’ve used high accuracy ground control (total station or RTK GPS rover for example) or have access to an RTK drone then what you’ve got is pretty good.

Just thinking about it a bit more there could be ways of ‘gaming’ the software to get higher accuracies. In other photogrammetry software its possible to add in ‘scale constraints’, essentially precise known measurements.

Whilst ODM & WebODM doesnt explicity support this, perhaps you could use the GCP system to apply local coordiante system. For instance you could use your UTM coordiante system and calculate out GCPs based on a base point at (1000,1000,100), then use your knowledge of the base line distances to calculate out the others and mark it up in the GCP tool.

Another way of doing the same would be using QGIS’s excellent georeferencer tool, where you could do as above, or reference to a Google Earth or Bing Maps background. Food for thought!