[TUTORIAL / TIP ] Alternative to cross section tool

I recently posted about the cross section tool, has very big limitations and is only usable for perfectly straight elements lik a bridge.

https://communities.bentley.com/products/3d_imaging_and_point_cloud_software/f/contextcapture-descartes-pointools-forum/147179/model-by-section-needs-improvement

I've came across an approach to workaround the limitations of that tool.

If our dataset has vegetation you will probably whant to run classification first so you run following steps on a ground+road visible points dataset.

You will first need  to extract road axis or a line that is paralell to the axis. In a 2 lane road that is pretty easy with the follow line tool. I'm not detailing this step here.

You could alternativelly drape a cartography axis, cleanup, facet, convert to spline and facet back.

We've the centerline extracted so we draw at start of it a line, perpendicular to it, this line will have the span of the cross section we shant to have, center it to the axis line.

Then using the array tool we will clone this section by distance, for this example I did clone each 5m. Enable "rotate" checkbox, this will orient cross section with road axis.

The array follow the axis in z too. This is important for later on.

You can adjust the lenght of specific cross sections to cover aditional cloud areas if needed.

In the Descartes attach cloud panel, there is a tool called "Drape element" This will project a line over the point cloud. Dissable generalization and try first the parameters with a couple of lines. There is a height parameters that seems to define the distance within the line and cloud are so if line to drape was in 0 and cloud at +50 and distance 10 (default) wont project, but as the array places lines following the axis, default value works just fine.  After projecting all the cross section at once (takes 2-3 mins) you may see  some spikes in the data, in my case the biggest ones where at the very center of the cross section, after a cleanup of 5 mins remove the biggest spikes.

Some terrible spikes in the centerline:

You may at this point extract cross sections by selecting one of theese or extract linear features following cross sections.

It does also help in understanding the cloud details, as many times specially in the sides where density is lower understanding shapes needs for the camera to orbit.

I'll try to fine tune params as there are some IMO unjustified spikes. This is why it is not a good idea to run line decimation when projecting as it will decimate over a non perfectly adjusted line. To avoid those spike what I would do internally is to mesh the cloud with Delaunay XY plane like in Cloud Compare. And project over that mesh but from the bottom, not the top.  Even I did use view projection from bottom theese spike presist. But with the meshing those should be avoided. I would import that mesh into MS and use the legay mesh line drape but MS does not handle well hi-poly meshes :\ It does a terrible job in fact.

Hope you all enjoyed and that is is usefull for you.

[EDIT] checking on teh side spikes issue. Theese are due to vegetation still present. Tested the heigh param aand is the distance from the line to che cloud that will be considered for projection, so if you've vegetation avobe it will pick those points. 

Original images (I uploaded instead because I may delete drive or whatever)https://1drv.ms/f/s!AuliwwO_siw1i0BDDCFiuvcY2041