This is a python implementation of the source code associated with (https://doi.org/10.5201/ipol.2017.208):

Christoph Dalitz, Tilman Schramke, and Manuel Jeltsch, Iterative Hough Transform for Line Detection in 3D Point Clouds, Image Processing On Line, 7 (2017), pp. 184–196. https://doi.org/10.5201/ipol.2017.208

The exact performance of the code has been modifed to be application specific and NOT exactly replicate the behavior of the source code.

---

To-do?:

• Output points in lines (either ID's or lists of points or ? maybe some intergration with flow is possible?)
• Output confidence values? i.e. eigenvalues or LSQ-fit value to identify line strengths? (Note: this is obtainable with a description of a line and list of points, resource scope exceeded?)
• Add energy dependent functionalities? Not likely, but could have minEnergy? Again, this likely exceeds the scope of this resource.
• Optimize code by using list indices instead of lists.
  • Rather than initialize X and remove candidate points in Y from X, include a boolean flag in X and Y (X init to all 1 and Y to all 0) and perhaps make the flag a line index in Y (would instead init Y to all -1 and assign line index as they are found)?
  • ?

---

Overview

sphere.py Generates vectors in a manner to maximize the their uniform distribution about the surface of the unit sphere.

• Starts with 12 vertices and 20 triangles of a unit icosahedron
• All triangles are tesselated (1 triangle => 4 smaller triangles, defined by reference to 3 vertex indices), then vertices are normalized to put them on the unit sphere, repeat SphereGranularity times
• Degenerate vectors (v=-v) are removed
• A sphere.pkl file is used to store uniqueVectors, vectors, and triangles in a manner that automatically pulls from the best source. sphereVertices(...) can also take the argument saveToFile=False and will recompute sphere direction vectors procedurally instead. If encountering any issues, consider deleting sphere.pkl and rerunning to refactor this cache file.

hough3Dlines.py Main run code. Also includes orthogonal least-square fit procedure definition (orthogonal_LSQ). Details can be found in code/IPOL-article. (This section to be updated?)