3D Laser Scanning - Meshing Point Clouds in Cloud Compare

hi and welcome to another UQ 3d scanning tutorial today we're going to be looking at machine cific lis meshing in cloud compan and mesh lab so we're going to be taking a point cloud we generated using our three p16 laser scanner and taking it through to our mesh so first things first you want to load up your point cloud file that you're going to mesh this is one that we prepared earlier we've basically exported this out of cyclone as a PT X so to preserve point normals then we've cut down out of the irrelevant data cut out all the trees and then joined and subsample the data so it's been subsampled to eight millimeters we have used that to reduce our point load from somewhere in the vicinity of around 100 million down to about 11 million points you can quite easily use point clouds with high densities however because of what we're doing it just basically makes it much easier for us to manipulate and you do not need extremely high densities for your point clouds to mesh so as I mentioned nowhere in the scan we do currently have the point normal data already inside however for the purposes of this exercise I will show you how to get that point normal data as often that data can be lost if it's explored in the wrong file format or it's gone through the wrong file format and it's something good to know so to start I'll show you how to check that you have point normals in your scan so one of the easiest ways to do that is just to toggle on lighting effects as they were calculated using the point normals and that's very easy to toggle on and off another way of checking is converting the point normals to a scalar field and then it's just done through your edit menu so this scalar field enables you to have a quick look and an ALICE of your mesh normals and as you can see here there's a couple of areas that that can come through not quite correct this should be fine for the intents of our exercise however that can cause issues that you might be able to show you in a little bit so to start off with I'll just show you how to register the normals if you don't have those and we'll compare and contrast the normals that we've brought in through our PT X that'll be generated by the scanner and the normals that are generated within cloud compare so you to this you go tools normals compute so we're just going to duplicate the scan first just so we can compare these and turn off this initial scan so then we're good so we just go edit normals compute so there's three ways to compute normals you can see from this menu it's through plain height function and triangulation now generally speaking for buildings we do it via a height function that's just because it's quite good at capturing sharp edges and a lot of the maths work out quite quite well planar can work so reasonably well but it's not quite so good with the sharper corners and translations really great for curved objects are more elements with large numbers of curves but just for the moment we'll just use our height function here so just using this height function we're going to look at using now said playing for a preferred orientation this just tends to work the best for large-scale scans the next thing we're gonna look at is our radius here now the default is generally pretty good though you can change it if you want it's just talking about the amount of area it uses to capture the surrounding points to help work out the normal so four reasonably dense point clouds like our own or point clouds with decent fidelity it's quite good but saved quite a low fidelity point cloud you might want to expand that out to just just capture as many points in order to work out that plane for that normal so I'm feeling pretty good about all this and we'll just press ok and then ouch start processing so now that's done we'll just take that back to a scalar field so that again is edit normals put the scalar field just change it down there so make sure we put the right scalar field selected for visualization and there you go as you can see that's much more kind of consistent in planar and it's looking quite good so what we're going to do is now compare the two scans and through a mesh and see how well they mesh so if you look up at this item here on the toolbar this is our posse in mesh button this is a plugin for cloud compare but this is how most meshes for point clouds are handled it's also available in the plugins menu as we can see here we've got up various sub menu items we want to set an octree depth of approximately 11 and we can have looked at our samples per node we've got a quite a good point clouds so we can leave that all pretty much how it is and last thing we just want to output our density as a scalar field this will enable us is our scalar field functions to cut out all the extraneous data that this meshing exercise will generate you'll see in a moment so we also get this option to import our cloud colors so we'll just select ok since this is a colored point cloud if not and that would be fine just to go now okay as you can see that's coming through we'll just turn off a point cloud here and just try and highlight our mesh out so we'll turn on our lighting here so let's just display turn on sunlight should give us another look at it just have a scalar field now and we'll just start using this scalar field to cut out some leaves low density points here we go this is start to give us a better idea of how it's looking we'll switch set back over to RGB that's not come through amazingly well unfortunately you see we've got a lot of um almost bubbling and where the normals are played with so now we're just going to try our other mesh settings so it's a similar process just using pressing that pacient mesh button it's all the same settings we'll just press ok okay so now let's come through and that's looking quite nice so what we're going to do is just cut out our additional data by density and you can see how quickly see quite quickly how useful that is that really enables us to cut out all this extra generous data that the meshing process creates so to actually cut that out we go to users by edit scalar fields filter by value and that brings in our points that we've registered on through that graph there automatically so that will cut out all that data automatically for us which will enable us to reduce that down a lot you see here that's all come through quite nicely we've got a new graph there with new densities and that means we've kind of cut all that additional data out for us so this is as far as we can go and cloud compare our next step is in through to our mesh lab and we'll repeat a lot of this process using the mesh lab processes but as you see it's come through very very nicely and that's quite successful mesh so we're just going to save this as a peel Y and just call it tutorial great chord and just savor it as a binary format and here we go all right we'll load up my shop now

Channel: UQ Architecture

Views: 79,259

Keywords: 3D Laser Scanning, UQ Architecture, Cloud Compare, Meshing, Point Clouds, Poission, Point Normals

Id: MS3Krxcy2j0

Length: 11min 35sec (695 seconds)

Published: Tue Mar 08 2016