Quad-remesh behaves inaccurately


I’m using some GIS data to create a wider scope of my location - terrain mesh.
Also, I have a survey done for the narrower scope around the parcel itself.
I’m combining those point-clouds to generate the terrain.

Firstly, I’m passing the point-cloud through the Delaney mesh component.
Then I’m rebuilding the mesh to correct for some overlapping points, and lastly, I’m reworking the mesh through the Quad-remesh component to get a cleaner mesh.

The problem I’m having is happening when Quad-Remeshing. While processing the points, the triangulated meshes follow the point cloud, but when transforming into a Quad-mesh, the mesh - sort-of “shrinks”.

Note the yellow points - representing the footing of a water-reservoir on the top of the hill.
The quad-mesh rebuilds below the point height.

Rebuild triangulated mesh:


How can i control the tolerance of the quad-mesh rebuild? Something wrong with the settings. I’ve tried it all.
Or maybe I should approach this differently?!

I’m questioning all the times I have used the quad-remesh component.

It is meant to be inaccurate.
But if you share the file maybe we can find a better solution.

And please tell what is your purpose of doing quad remesh

What do you mean “It’s meant to be inaccurate”?! That’s absurd.

I’m not asking for a solution to my script.
I’m just asking for someone, better versed in the matter, to explain to me - how to approach the quad-remeshing protocol - to follow the vertex constraints of the input mesh.
Any mesh.

Is the model at state plane coordinates? If the model is far away from the origin it is possible that the math becomes inaccurate due to the number of digits in the distances. Normally this might happen outside 8 digits left of the decimal point. What is a typical coordinate on that mesh in Rhino world space?

1 Like

Have you tried TriRemesh instead of Quadremesh?
I usually encounter the same problem if I try to rebuild a mesh by quadremesh but do so with too little target count for the geometry.

you have set it to 150’000, but I can’t tell wether its not enough for your purpose because I can’t tell the extend of your pt cloud.Try and set it higher.

1 Like

It’s maybe expressed in a unlucky manner, but @Quan_Li is not entirely wrong with his statement. you cannot expect vertices to be placed in a regular manner when keeping the existing vertices to be kept the same. in this case you’d rather use -_reduceMesh
or said differently, spikes to be flattened out and have perfect accuracy. If you alter the accuracy, the spikes will be less flattened and if you flatten more, more information is lost.

1 Like

Increasing the Target Count is your only chance here.
Preserve hard edge will not work, because you don’t have it.
You don’t have a guide curve, so Guide Curve is irrelevant in your case.

If you don’t understand what is guide curve doing, you can see my other post using it.
The purpose of QuadRemesh is for better topology, not for accuracy, but increasing the quad number will get you there with the cost of heavy computing.

1 Like

Yes, it is.
I thought about that. Moving the World Origin doesn’t help.
I’m building the terrain data at the state coordinates so I can transfer some back into QGIS - “simulating” georeferencing.

There are two types of coordinates:
7633809.515825521;4600036.281994683;265 < [height map data] and 7633719.79;4598832.38;329.55 < [survey-the yellow points]
Could this difference be a problem?
The projected difference between the point and mesh is over 2 meters.
The Scope is 3x3km

On the other hand - I’m thinking @Quan_Li and @benedict are right > If I invest in the heavy computation, I could get the mesh “there”.
I was expecting the remeshing to take the vertices of the previous mesh, but, in retrospect, my thinking was completely wrong.

I should start using the Tri-mesh calc… but the Quad-mesh looks soooo pretty. :sweat_smile:
The quad-mesh is easier to manipulate in model, if it’s needed.

For the architecture part, I’m taking out a small chink of the terrain and sending it to World Origin via referenced point, so that I can move it back to correct position.

I’m thinking of giving that chunk of points more detail while quad-meshing to get it more accurate.

Thank you guys :people_hugging: