I’ve spent some time playing with webODM using docker and had some good results, for a variety of sites. Although my aging laptop struggled.
I’ve just got a new Dell XPS, and invested in the native windows install of webODM, to streamline things somewhat.
However, using default settings I discovered that DSM/DTM is not longer on by default, and the built in 3D viewer wasn’t displaying anything at all, despite a pointcloud appearing in the zip download. The fix I discovered is to reduce pc to medium, then it works. Scrap that, it’s now stopped displaying the pc that 10 mins ago was working fine
I’d like to try to partner with my former University and Google Summer Of Code again this summer to see if we can get accepted to work on Rolling Shutter Correction. We got passed by for this work this past Summer.
Thanks for the tip about zooming to full extent, has worked a treat! Good luck with the summer of code, I’ll try and keep an eye open for if you need any supporter emails or such like. I’ve been trying to work out if it would be possible to somehow distort the images based on the vertical compression that the rolling shutter causes, although this would presumably alter the location of the point of the image. Maybe that distortion could even be generated by altering the vertical dimension of each pixel based on the velocity and sensor readout time. Thanks for your help anyway
In some cases the correction is even more complicated- such as when taking images on the timer, moving to the next waypoint at the end of a run when the drone is simultaneously moving and rotating. I’m not sure there is enough information saved to be able to correct for that.
I almost exclusively use the intervalometer to take the photos. I assume that pix4d for example doesn’t have that to deal with only taking photos at specified locations once the drone has straightened up. Usually I delete the images where the drone is turning, so that wouldn’t be such an issue. I am now tailoring flight speed to the site, trying to fly the site as slow as possible to reduce possible errors (which is fine on small sites). For me, if a small site takes 20min of flying instead of 10min it makes little difference and halves distortion, but that’s not practical on bigger sites, and the intervalometer doesn’t allow shoot-move-shoot which is an alternative.