Program to reconstruct sorghum plants from depth images and make measurements of the reconstruction

This software is provided as is, without warranty or guarantee that it will work, and many features remain untested.

• A series of depth images representing a single plant are converted to point clouds and registered.
• The pot is segmented out from the registered clouds, and a final round of registration is performed.
• The registered cloud is meshed (using a combination of MeshLab and Screened Poisson Surface Reconstruction; these are external to the SorghumReconstructionAndPhenotyping program).
  • MeshLab - http://meshlab.sourceforge.net/
  • Screened Poisson Surface Reconstruction - http://www.cs.jhu.edu/~misha/Code/PoissonRecon/Version8.0/
• The plant mesh is segmented into stem, leaves, and inflorescence using a machine learning approach (trained on point features)
  • MultiBoost - http://www.multiboost.org/
• Individual leaves are segmented out using supervoxels and geodesic lengths across the mesh.
• Measurements of the entire plant mesh and individual plant organs are made.

Ryan McCormick at ryanabashbash@tamu.edu or http://codextechnicanum.blogspot.com/

• RapidXML - http://rapidxml.sourceforge.net/
• OpenCV - http://opencv.org/
• PCL - http://pointclouds.org/
  • (and the PCL's dependencies; e.g. Boost, Eigen, VTK)
• Intel's Threading Building Blocks - https://www.threadingbuildingblocks.org/

First, make sure you can compile a project that depends on PCL and OpenCV (e.g., http://codextechnicanum.blogspot.com/2014/10/installing-point-cloud-library-and.html). On an Ubuntu system, PCL and OpenCV can be obtained from the apt package manager, but I recommend compiling PCL from source (e.g., http://pointclouds.org/documentation/tutorials/compiling_pcl_posix.php) so that you have access to an up to date version and ability to turn on things like fitting NURBS and B-splines (e.g., http://pointclouds.org/documentation/tutorials/bspline_fitting.php).

If you can compile PCL and code that depends on PCL, you should be able to use those build options in your IDE of choice, add all of the source files to your project, and compile. At some point in the future, an improved installation process will hopefully be added.

As a personal note, the hardest part (for me) was tracking down where on your system the necessary libraries are stored. You can find a list of the necessary includes and libs to link to in the MAKEFILE. The MAKEFILE itself was autogenerated from a CodeBlocks configuration and is untested, and is provided more as an example of the includes and libs you need rather than something that you can expect to run make with.