In this episode of Lightning GIS we will demonstrate the procedure for filtering noisy data due to reprojection error out of a sparse point cloud in Agisoft Metashape with the gradual selection tool.
Hi, Great video. I am very new to metashape, so am trying to learn it. Could you explain why all the cameras are deselected before you run the reprojection filter? Doesn't doing the reprojection adjust camera locations? If the cameras are turned off what is changed? I watched another of your videos on the other two gradual filter options (Reconstruction uncertainy and projection accuracy) and you have all cameras selected for running those filters. Any help appreciated.
Hello, can you briefly explain why only optimizing with f, k1-k3, cx,cy, p1, p2 for Reconstruction Uncertainty and Projection Accuracy? Then later on after adding markers (or not … if ppk reference positions), when you optimize to all parameters - any reason not to also do 'fit additional corrections'? What is the additional corrections anyways, do you know? During alignment, is it appropriate to select adaptive camera model fitting? Even though after will be doing filtering and optimization the way you show. i am just trying to get some sense of WHY for these steps you mention.
My workflow has evolved a little since I produced this video (in part due to further experience and experimentation and in part due to updates to the software). I now use “adaptive camera model fitting” and “fit additional corrections” and I don’t bother with the individual check boxes. I can’t explain the different individual parameters (at least off the top of my head 😃) but there are articles and websites that do and are easy to find.
Great Video. It would be great if you can make more videos on Agisoft Metashape. Can you let us know how to remove distortion in Orthomosaic. Even following all the steps I have got distortion in building edges, if you can post a video on the same would be great.
What if, during a culling of points, you get the message that some cameras have insufficient whatever and will be reset? What does that imply? Is there a procedure for that scenario?
In that case the gradual selection has removed all the tie points on that photo have been deleted and it can no longer be aligned to other photos. Usually this is not a big deal if you have sufficient overlap and that the photos lost are not all clumped together (in which case you will be left with a hole in your model). This implies that the photo that was lost had minimal tie points and the ones it did were very noisy. Probably not a big loss. Losing a photo here or there isn’t usually a big deal.
@@lightninggis4758 Thank you for clarifying (and responding quickly!). One other question I had was your gradual selection order of the last two operations - I've seen those two flipped on the Geospatial Tips channel here: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-je79gV8HsZI.html Do you have a rationale for one or the other?
