Library for 3D LiDAR perception

Author: Alastair Quadros

LaserLib is a c++ library containing work from my PhD thesis on processing 3D point clouds from a Velodyne LiDAR for object classification.

It is developed for close use with python and numpy: all functions have a python wrapper.

It consists of:

• Datastructures for selecting regions of points on range images / point clouds
• Surface normal routines
• Features such as PCA, spin images, line images, interfaces to PCL features
• Knn classification
• Affinity propagation clustering

The best documentation is in the python interfaces, which largely mirror the c++ ones.

http://www-personal.acfr.usyd.edu.au/a.quadros/LaserPy/index.html

• Eigen 3 http://eigen.tuxfamily.org/
• Boost

Optional, recommended:

• Python (>2.7)
• Flann http://www.cs.ubc.ca/research/flann/

Optional:

• PCL (Point Cloud Library) http://www.pointclouds.org/

@CONFERENCE{quadros2012feature,
author = {Quadros, A. and Underwood, J.P. and Douillard, B.},
title = {An Occlusion-aware Feature for Range Images},
booktitle = {Robotics and Automation, 2012. ICRA'12. IEEE International Conference on},
year = {2012},
month = {May 14-18},
organization = {IEEE}
}

http://db.acfr.usyd.edu.au/download.php/Quadros2012ICRA_OcclusionAware.pdf?id=2522

@PHDTHESIS{quadros2014
author = {Quadros, A},
title = {Representing 3D shape in sparse range images for urban object classification},
year = {2014},
school = {The University of Sydney}
}

http://hdl.handle.net/2123/10515