PointCloud Skeletonization is a work-in-progress pipeline to read Point Cloud from a lazer scanner, register multiple viewpoints and extact the visual skeleton of the scanned objects. We are particularly interested in Corn plants in field condition. At the moment, the algorithm work reasonably well with fully scanned plant; however, missing parts (due to ooclusion) and non-linear skeleton body is still to be addressed.
| Input PointCloud | Previous algorithm | Enhanced algorithm |
|---|---|---|
 |
 |
 |
Based largely on the paper L1-medial skeleton of point cloud and their implementation, PointCloud Skeletonization make several changes to make the algorithm works better with plant:
- A new randomize algorithm: One critical step of skeletonization is sub-sampling from Point Cloud. This step reduces computational cost and prevent overfitting. However, a normal randomize algorithm doesn't sample the space evenly, instead we use quasi-random generator to generate low-discrepancy sequence of points, which help with finer detail.
- Various tweak to reduce aggressive merging of branches: Plants usually have much more branches compared to everyday ojects. Various steps of the algorithm try to smooth the skeleton points, which unintentionally merges real branches. We tweak its hyper parameter so the fitting process gives better result. This part is empirical
- Move windows specific code to cross platform Qt4 code and Cmake build system.
Install compiler and tooling for compiling software
sudo apt update
sudo apt install build-essentials
Install dependencies
sudo apt install cmake
sudo apt install libqt4-dev
sudo apt install libglew-dev
sudo apt install freeglut3-dev
Change the directory to root project folder
cd ./l1windows/Point-Cloud_Procesing_1_0/point_cloud_1_0_source_code/PointCloud
Create a folder for out of source build
mkdir build
cd build
To build on Unix OS (i.e Ubuntu, Mac OS)
cmake ..
make -j8
To build on Microsoft Windows
cmake .. -G "Visual Studio 10"
double click the generated project file -> click build
Please follow the tutorial videos, which are available in Tutorial folder for step by step guide to use this software.
- Cmake - CMake is an open-source, cross-platform family of tools designed to build, test and package software
- Qt Framework - Qt is a free and open-source widget toolkit for creating graphical user interfaces as well as cross-platform applications that run on various software and hardware platforms.
- OpenGL - Open Graphics Library (OpenGL) is a cross-language, cross-platform application programming interface (API) for rendering 2D and 3D vector graphics.
- Glew - The OpenGL Extension Wrangler Library (GLEW) is a cross-platform C/C++ library that helps in querying and loading OpenGL extensions.
- Glut - The OpenGL Utility Toolkit (GLUT) is a library of utilities for OpenGL programs, which primarily perform system-level I/O with the host operating system
We follow Iowa State University authorship policy, please send requests to us at the address: baskarg@iastate.edu
We use git for versioning. For the versions available, use the git tagging.
- Truong Tran - LinkedIn
See also the research project info at lab website for detail information.
Copyright 2018 Baskar Ganapathysubramanian, all rights reserved
We gratefully acknowledge support from Plant Science Institute at Iowa State University.