Terrain Reconstruction from Point Clouds Using Implicit Deformable Model

• Python package, implementation of Morel et al. 2020.

Setup

Requirements

• Linux (tested on Ubuntu 21.10)
• PCL 1.11
• Python 3.9.7
• GNU Scientific Library

Build

Compile the package by running the following commands:

cmake .
make

Usage

Import the package:

import dtm

Build the implicit surface model:

• MINIMUM_POINTS: the path to the ASCII file containing the minimum points to approximate.
• BB_POINTS: the path to the ASCII file containing the 4 points of the minimum bounding rectangle of the minimum points.
• MINIMUM_POINTS_NUMBER: the minimum number of points in the quadtree leaves.
• MINIMUM_SIZE: the minimum size (in meter) of the quadtree leaves.

surface = dtm.dtm(MINIMUM_POINTS,BB_POINTS,MINIMUM_POINTS_NUMBER,MINIMUM_SIZE)

Apply the deformation to the initial implicit surface (optional):

• NUMBER_ITERATIONS: the number of iterations of the deformable model.
• GAMMA: the ratio adjusting the elasticity of the surface.

surface.applyDeformableModel(NUMBER_ITERATIONS,GAMMA)

Extract the surface mesh model by polygonization of the implicit surface:

• NX, NY and NZ: the division of the polygonizer 3D grid

surface.polygonize(NX,NY,NZ)

Save the surface mesh model as an .obj:

• OBJ_FILE: the path to the .obj file.

surface.exportDTM(OBJ_FILE)

Plot the results of each steps of the algorithm (optional):

surface.display()

Example

To run the algorithm on the testing data, simply run:

python test.py

The graphical output should be similar to the image below