Generate stl mesh from csv

Question

I have a csv file that encodes elevation data of a 9000x5000 resolution photo of a dam, and I want to generate an stl file that I can print on my 3d printer. The csv has 3 parameters: X, Y, and Z, where X & Y denote a pixel location on the photo and Z is the elevation of the topography at Pixel (X,Y). I want a 3D stl file of the photo in question with the nooks, crannies, and hills denoted by the elevation data. I attempted to do this using the Python library numpy-stl, but I'm realizing that I just don't know enough about stl files to accomplish the print job. Any information on how I can make this file, or on how stl files encode information would be remarkably helpful.

Answer 1

What you have is a collection of vertices. However, what STL encodes is a list of triangles. So what you need to do is to construct triangular mesh connecting your vertices. I assume that you want to code it yourself and I will provide you some hints on how to do it.

First, you should find out, whether your points are structured or not. By structured I mean that the x- and y-coordinates of the points form a regular grid (this is often the case when the data are obtained from airborne scanning).

If the points are structured

If the points are structured, you have a decent chance of constructing the mesh by yourself. Denote P(i,j) the point (X[i],Y[j],Z[i,j]), that is, the point whose first coordinate is the i-th number in the list of the x-coordinates, whose second coordinate is the j-th number in the list of y-coordinates and whose third coordinate is the z-coordinate corresponding to the first two. Figure 1 illustrates.

Now you need to decide the orientation of your triangles. They all should be either positively (counter-clockwise) or negatively oriented (clockwise). Assuming you decide for negative orientation, you can now create the triangle connecting the points P(i,j), P(i+1,j) and P(i+1,j+1). See Figure 2.

In the next step, you can create the neighbouring triangle with the same orientation, say P(i,j), P(i+1,j+1), P(i,j+1). See the darker triangle in Figure 3.

Proceeding the same way for all suitable i's and j's, you can obtain all the necessary triangles.

The only missing part is the normal for each triangle. This you can compute with a cross-product as suggested in this answer.

Now you are ready to write everything to your file. Following the description on Wikipedia, for each triangle you write its normal and the three vertices.

If the points aren't structured

In this situation you don't have much chance of creating your mesh without a significant study of meshing methods (you will probably need to learn about Delaunay triangulations). Fortunately, there are some libraries available. For instance, Fade2.5D is a C++ library that can help you and it is free of charge for personal non-commercial research. Maybe there is something similar available for Python as well.