Hi everyone, This is my first question so I hope I am posting this in the right place. My goal is to capture roads with the intent of animating vehicles, crashes, etc on the meshed road surface. I have issues like this with a drone as well but images for this project were taken with a DSLR camera on a pole. Reality Capture aligned all the inputs after I found a helpful tutorial on alignment settings. At first glance, the project looks great but closer inspection shows some areas for improvement. Here is the textured model. Looks pretty great to me.
Looking at Solid Mode though.
I hope you can see the large sections of road that are slightly elevated or sunken in from the rest. I have tried simplifying which just makes the edges sharper, and smoothing which seems to have no effect.
I want the road surface to be flat as it is in real life because animated vehicles will react to the bumps in the road.
How should I attack this problem? In alignment? Editing in something like Blender? Using the RC edit tools?
These slight elevations are a result of misalignment since to limit misalignment it is recommended to have cameras oriented toward the object of interest, and to have images taken from various distances from the object. In this case, it would mean that cameras should be oriented toward the ground as much as possible, and two layers of image sets would be enough (the height you used already, and one from the double that height).
To eliminate those bumps you can use the smoothing tool or simplifying tool but remember that those two tools can work on selections. That means that you can use the lasso selection tool (or any other selection tool) and select the part which you want to smooth/simplify. Also, it may take some adjustments to the settings of those tools to get the preferred results.
You can also export your model, and edit it in software like Blender. If you plan to re-import it into RealityCapture, make sure to not change the position and orientation of your model!
What drone have you used? If your drone lacks mechanical shutter, this might be your issue. You can try to use Brown3+t2 distortion model in alignment settings.
I normally use the Phantom 4 Pro. Pretty sure it has a mechanical shutter. In this case, as an experiment, I used a Canon T7 on a tall pole. So it has a mechanical shutter as well. Thank you for the input though.
I work in forensic engineering and would like to have smooth surfaces for vehicles as well. I find that oftentimes a problem is road striping or shadows appear raised or sunken from the rest of the road surface
So this particular issue was caused because I used a DSLR camera on a tall pole (20-25 feet tall). I used a 10mm lens but have an APSC sensor camera. So the effective focal length is 15 to 16 mm. I think my trouble came about because I couldn’t capture enough of the environment in each individual shot. Still got a decent result but I wanted better. I normally use a Phantom 4 Pro and I take Nadir (straight down) shots in a grid pattern. Then take oblique shots circling the area at 2 or 3 altitudes. Using all these images provides the best results for me. A key, and I am still working on this, is to decimate the resulting mesh into one with large enough, few enough polys to be usable. I have also gotten excellent results combining laser scans with the drone photogrammetry.
Maybe you know all this already and I am going down the wrong path for you. Hope this helps.