3
Rant: Our drone mapping software gave us totally wrong elevation data on a job in Denver last week
I'm still fired up about this. We were using a pretty popular photogrammetry platform for a cut and fill estimate on a 2 acre lot near downtown Denver. The software said we had a 12 foot elevation change across the site, which sounded reasonable. So we bid the job based on that. Then we got a survey crew out there to double check before we broke ground, and you know what? The actual change was only 4 feet. That's a massive difference in material costs (we were off by like 30 truckloads of fill dirt). I think the issue is these drone programs get thrown off by trees and buildings casting weird shadows. The guy running the survey even showed me how the point cloud had these phantom high spots from a nearby apartment building. Has anyone else had their drone mapping software totally miss the mark like this, or did we just pick a bad program?
2 comments
Log in to join the discussion
Log In2 Comments
king.aaron18d ago
Jumping off what you said about the point cloud - we had the exact same phantom high spot issue on a job near a big reflective warehouse. The software kept seeing a false ridge that was just the sun bouncing off the metal roof. Ended up having to re-fly the whole site at noon with overcast skies (which is a pain in Denver, you know how that is) just to get clean data. It's wild how much the sun angle and nearby building reflections can throw off the ground detection, especially with photogrammetry vs lidar. I bet your program was trying to make sense of the apartment building's shadows and just guessed wrong on the real ground level underneath.
8
victorbutler18d ago
Three flights over three different days finally fixed it for us. Turns out the ground detection algorithm was locking onto the apartment roof's thermal signature in the morning light. We switched to flying right when the sun was highest and the shadows were shortest, and the point cloud came back clean finally.
6