Agisoft Metashape
Agisoft Metashape => General => Topic started by: smescarzaga on September 17, 2018, 06:04:00 AM
-
Because I'm working with coastal imagery, nearly every photo in the dataset (~4,000) has water pixels in it. This has caused a lot of noise in the dense cloud, particularly near the land-water interface (which is were I'm interested in studying) I think it'd be rather inefficient to try and mask every photo manually in Photoscan or photoshop. I had a tip from someone to, during the gradual selection phase, remove all tie-points that had an image count of 2. Not only did this remove much more than half of the points (not the best overlap), but I still ended up with a bunch of noisy points in the water. T
Since the ROI is rather large, my next thought was to try filtering these points out systematically with a filter (maybe in cloud compare) looking for areas of high frequency z change.
Has anyone had any luck with filters in or outside of cloud compare filtering out noise due to water? I've attached a screenshot of some of the noise.
Thanks.
-
I can suggest you a fast procedure to mask water if this is what you want.
Draw a polygon enclosing water on the dense cloud, and keep this shape selected. (Try to follow the very border of the water so the polygon vertices have right elevation)
goto tools, dense cloud select by shape.
assign water class to the selected points (right click on pink area)
Build heightfield mesh (high poly number does not take longer than low poly) -advanced dropdown and use class watter only, and disable interpolation
goto file import-masks-from model and apply to all photos...
But there is a much better way to obtain a clean water surface. After having assigned water class to points, simpy build the DEM file excluding water class, my friend. That works just perfect!
-
Awesome, I'll definitely give this a go. Many thanks for your reply here and to my other thread.