Point Cloud Registration

Point cloud registration is the process of aligning 3D point cloud datasets to eachother when they contain overlapping regions. This repository contain an implmentation of such an algorithm which is the main focus for my master thesis. The thesis is included in the folder as well, describing the implementations and theoretical derivation of the algorithm.
Most of the work is based on articles by Qian-Yi Zhou, Jaesik Park and Vladlen Koltun, who are the authors of the Open3D library.

Folder structure

The repository contain two folders,

• PointCloudRegistration
  • Contain all files needed for compiling the programs Registration and GenerateData. A visual studio project is included and a readme describing the functionality of the programs in depth.
• Testing
  • Containg shell scripts and matlab code used for testing and visualizing the functionality of the program.
  • Technical report describing the process of developing the code and the theory the behind problem and method.

Prerequisites and libraries

In order to compile and run the code the following libraries must be available: (the versions specified in paranthesis are the tested versions).

• Eigen (3.3.5)
• Open3D (0.4.0)

Installation

Visual Studio 2017 projects are available in the Point Cloud Registration folder along with a solution file (.sln). These should be updated to include the correct directories for the libraries. Makefiles are also included for compilation of the programs. Again some alteration will be required in order to locate the correct libraries.

Running the tests

For direct usage of the program files in the commandline please see the readmes located in the PointCloudRegistration folder. These describes the functionality in depth. The test scripts located in Testing/shell is designed to run through the main test script runtest.sh. Runtest.sh is designed to setup temporary file locations and log files for the tests. Additional information can be found in the README located in the testing folder.
Matlab version R2018b is used for testing.

Authors

• Tim Felle Olsen - TimFelle

Project supervisors

• J. Andreas Bærentzen