Hey, I’m running into an issue with RealityScan where my dataset aligns into one component, but the result is way off. I captured the site using a drone with both orbit and straight-line passes, and used the video sequence tool to extract frames.
After alignment, the model looks like the same area has been duplicated and slightly shifted — both vertically and horizontally. It’s like the same structures show up multiple times but in the wrong spots. The camera tracks also look split and drifted.
I tried adding 4 control points across the loops and re-aligning, but it didn’t seem to help. I’m not getting multiple components, just one — but it’s internally misaligned.
Has anyone run into this? Is it a control point issue, or maybe something with the video frames or too much overlap? Would love any tips on how to clean this up or prevent it in the future.
Hi thanks again for the tips, really appreciate it!
I’ve followed your advice and now have 15 control points around the dataset, each linked to about 20 images. Each point is visible from different angles — captured during two linear passes in opposite directions and also from orbit passes, so they’re seen from four directions.
Unfortunately, I’m still seeing the same alignment issues. The model doesn’t show any banana effect — no bending or warping — but there’s misalignment.
Would adding even more control points help at this stage, or is it more likely an issue with the capture method?
Is there a way to share the project for deeper review? I’d really like to get this working and understand how to avoid it next time.
