Mixing Drone and DSLR images getting two components, can't combine even with control points?

Hello

I’m pretty new to this Photogrammetry game but really enjoying the science behind it.

I thought I would start out small and modelled a porcelain child’s lamp quite successfully and then I thought I would have a mini practice on what I really wanted to use Reality Capture for and that’s outside scans of buildings and land marks.

  • So I thought I start small and just set up 2 lines of 3 chairs in my backyard.
  • Then using a DJI Mavic Mini Done I ran a straight down view over the top taking about 45 images
  • Then using the Drone I did a rough circle around the chairs at about waist height
  • Then lower again
  • Then I got out my Olympus EM-1 Mk2 with 24mm equivalent (same roughly as the drone) an took a heap of ground level and knee height shots.

 

My problem is whenever I try and align the images, I get a great mix of drone and DSLR shots covering the scene in one component of about 208 images as here…

 

but I get another component of 45 shots consisting solely of the straight down drone shots and excluding everything else. What’s more that component is at 45 degrees to the other one….

 

This is the best alignment I can get using Max Features per image 80000 and Preselector Features 40000, Low Overlap and Ultra Sensitivity (anything other than ultra is worse i.e has more components)

 

If I drop in Control Points I can get one component but for reasons I can’t explain the 45 degree tilt of the straight down Drone shots is retained turning it into one big mixed up mess!

 

I’m trying to treat this as a one off thing where I can’t go back and retake or add photos. Is there any suggestions as to how I can get these two to join seamlessly?  

 

Also I notice all the drone shots have residuals is this part of my problem?

 

Any help grateful received….

 

Many thanks

Andrew

 

The residuals should not be an issue, the angles and the overlap is. You can’t go from top down to 45 degrees immediately. There needs to be better angle work in-between so there is no big change in what the picture sees continuously.

Thanks Erik

This is good to know. So for something like thus (assuming you could scale it up to something much larger) when doing those overhead drone shots are you better having the drone gimbal at say 70 degrees (as opposed) to 90 degrees straight down, and going a bit wider? I’m thinking doing that you are halfway to having a closer angle to the next row of circling drone shots? 

Also am I correct in thinking control points won’t let you stitch “anything” from a scene there needs to be some fundamental connecting “flow” between images? 

Thanks for your comments

 

Well I’d go for 90 degrees then 70, 45 and then side shots. The overlap should be like 60% so the camera position motion has to be really fluent, just as angles.
Correct, to connect some components via CPs there has to be something matching at least. It can’t look too different.

Cheers Erik this should be easy to do once I start running mapping missions and can just run multiple passes with different gimbal tilts. I did do another trial today with just 60% gimbal tilt and was much better but your suggestion would totally nail it.

Many thanks!

Glad to hear I helped, good luck with the projects!