I am beginning some work with LiDAR sensors (mounted on small unmanned aircraft systems), and would like to know if there is a way to transform LAS data to a solid, 3D model that can be used for object-based image classification? Specifically, I am looking at LiDAR sensors for vegetation mapping and would like to be able to classify vegetation within the 3D model. Is there a way to get to this point using ArcGIS Pro? I've used ArcGIS Pro for remote sensing projects, so I'm familiar with the program. Do I need to convert the LAS file to multipoints, and then krig them into a 3D model?
Any advice would help, thanks!
I'd need to get some more specifics on what about the vegetation you are trying to classify or cluster together. And I assume that by "solid 3D model," you essentially mean stitching together the LAS points to create surfaces that connect each point. That is something I've never heard of being done for vegetation and is probably unrealistic for any sort of densely vegetated areas. Extremely dense UAV collected point clouds could make it more plausible, but understory vegetation is notoriously difficult to characterize, let alone represent as a surface.
The primary OBIA that is performed with lidar data is tree segmentation. That is, identifying individual trees or "tree approximate objects" (TAO) from a canopy height raster or directly from the point cloud. We say TAO because it is extremely difficult to accurately delineate tree crowns, particularly in dense forests, which means that some of those segments will actually represent 2 or 3 crowns.