Plane segmentation research

Description

This repository contains materials for researching semantic and instance plane segmentation from point clouds collected from LiDARs.

Content

• src
  • algorithmsForPointCloud - this folder contains files with methods that allow you to transform point clouds, extract some information from them and label files.
  • open3DTool - this folder contains an MVP application containing: presegmentation by planar labels, segmenting planes by n points.
• notebooks
  • About_segment_plane - some information about segment_plane method from the Open3D library
  • Some_functions - some set of functions for working with point clouds
  • LeastSquaresFitting - implementation of linear regression and SVD-fitting methods to solve an optimization problem, as well as a few examples.

Running unit tests

1. Installing packages for testing

   python3 -m pip install -r requirements.txt

2. Running unit tests

   python3 ./scripts/run_tests.py

Installing packages for src folders

1. Go to the required directory

cd src/algorithmsForPointCloud
or
cd src/open3DTool

2. Install required packages

python3 -m pip install -r requirements.txt

3. A detailed README can be found in each of the directories

Example of running Open3DTool

python -m src --path_to_bin_file data/000.bin \
        --path_to_save_label_file data/labelFile.pcd.labels \
        --path_to_save_object_file data/labelFile.pcd.objects \
        --path_to_pcd_file data/000.pcd \ 
        --distance_to_plane 0.06 \
        --count_points_to_pick 5