matlab-scanmesh-toolbox

Overview

This is a toolbox of matlab functions to import, visulize, process, and export 3D meshes.

Primary appliation is for processing scanned 3D meshes. Our lab generates
3D reconstructions using photogrammetry software with extremely high resolution input images.
These meshes often have millions of vertices, are incorrectly scaled, and display hard-to-find
artifacts. This toolbox provides methods to help deal with these side-effects.

All of the provided methods run “fast” on very large mesh files. Parallelization is implemented using
default matlab functions whenever possible.

Methods

Import Export

read_object, write_object

Methods to read and write .obj files. These files will parse vertices, vertex normals, vertex textures,
faces, face normals, and face textures. Does not save any material or texture files. As far as I know
this is the fastest existing .obj reader implementation for matlab.

Object Visualization

vis_object, vis_object_segment

Methods to visulize objects using matlab’s patch function. vis_object displays the entire object in
a single color, while vis_object_segment will highlight a specific part of the object.

Camera and Figure Tools

vis_copy_camera, vis_paste_camera, vis_fill_camera, vis_change_background

Copy and paste are methods to save and apply the current camera orientation and target, as well as xyz
limits. Fill camera expands or shrinks the xyz limits of the figure to the edges of the figure so that
the object is easier to see. Change background removes the axis from a figure and replaces the background
with a solid specified color.

Colorize

colorize_recolor, colorize_curvature, colorize_centroiddistance

Recolor will recolor an input object to the specified color. Colorize curvature uses a simple approach to
color vertices based on the placement of their adjacencies. Colorize centroid distance color vertices
based on their distance from the object’s center of mass.

Rescale Object with Known Ground-Truth

perform_manual_rescale

Method to scale an object using known ground truth object. Requires an object of known
physical size which was scanned at the same time as the object to be scaled (I recommend a cube).

Align Two Objects with ICP

perform_manual_align

Method to align two objects to one another. Requires selecting three points on each object. Object
alignment is updated after every point, and the final point snaps the objects using ICP.

Transform Object

perform_rotation, perform_scaling, perform_translation, perform_matrix_transform

Methods to rotate, scale, and translate an object about a given axis, point or with a given matrix.
Each transformation returns a matrix, and matrices can be applied in sequence using matrix_transform.

Recompute Object Normals

perform_recompute_facenormals

Method that recomputes an object’s vertex normals and face normals.

Delete Specific Vertices

perform_delete_vertices, perform_delete_duplicate_vertices, perform_delete_unreferenced_vertices

Methods that delete a list of vertices from an object, identify and remove duplicate vertices
from an object, or remove vertices that are not referenced in the face list. Cleans up faces and
and vertices after removal. Useful for removing identified artifacts or cleaning up an object.

Find Connected Components

compute_connectedcomponents

Method to devide an object into its connected components. Useful to identify small artifacts.

Find Specific Faces or Vertices

compute_find_faces, compute_find_vertices

Methods that return all faces that refernce a list of vertices, or returns all vertices that are
referenced by a list of faces.

Compute Edges and Adjacencies

compute_adjacencylist, compute_adjacencymatrix, compute_edges

Methods to compute the adjacencies of an object, or edges of an object in stable or sorted order.

Surface Pathfinding

astar, dijkstra

Given start and ending vertices, vertex list, and adjacency list, will return the shortest path between
two points. Implementation of dijkstra is essentially the same as astar with the cost function set to
increment by one for each edge. Boolean value within each function lets you visulize the process.