AABB Tree Demo

The AABB tree demo showcases several

algorithms where the AABB tree is put at
work with polyhedron facet and edge
primitives. It also provides a menu for
benchmarking so that a user can quickly
get an order of the performances on her
models for both intersection and distance
queries.

The application accepts as input (and

through drag & drop) OFF files which
represent triangle surface meshes. The
main window displays the input triangle
surface mesh in wireframe mode and
allows interaction through the QGLViewer
arcball library.
Inside Points

Menu Algorithms / inside points allows the

user generating a specified number of
points inside the input surface mesh
(which is assumed to be intersection free
and without boundary).

The demo constructs an AABB tree with

the mesh facets and generates many
intersection queries with randomly
generated rays (the source of each ray is
randomly chosen inside the mesh
bounding box and the direction is
arbitrarily chosen for all rays).

Each query counts the number of

intersection between the ray and the input
mesh in order to determine whether the
ray source point is inside or outside (inside
if odd, outside if even).
Boundary Segments

Menu Algorithms / boundary segments

allows the user generating a specified
number of horizontal slices from the input
surface mesh.

The demo constructs an AABB tree with

the mesh facets and generates many
intersection queries with a set of
horizontal planes. The objects returned
(generally segments) are depicted in
green.

The Algorithms / Edge points menu

constructs an AABB tree with the mesh
edge segments and generates random
plane queries. The objects returned
(generally points) are depicted in red.

The Algorithms / boundary points menu

constructs an AABB tree with the mesh
facets and generates random line queries.
The objects returned (generally points) are
depicted in red.
Signed Distance Function

Menu Algorithms / signed distance

function to facets allows the user
generating a cross section image
(100x100) of the signed distance function
to the input surface mesh facets.

The demo constructs an AABB tree with

the mesh facets and generates within the
cross section a number of distance and
intersection queries with points and rays.
As for the “inside points” menu each ray
query counts the number of intersection to
determine the sign of the distance (inside
versus inside) while each distance query
determines its magnitude.
Unsigned Distance Function

Menu Algorithms / unsigned distance

function to facets allows the user
generating a cross section image
(100x100) of the unsigned distance
function to the input surface mesh facets.

The demo constructs an AABB tree with

the mesh facets and generates within the
cross section a number of distance
queries with points.

Another menu provides the same

functionality but for the mesh edges.
Refinement

Menu Algorithms / Refine allows the user

refining the current mesh through
• recursive longest edge bisection until a
maximum specified edge length is
matched (shown). The goal is to check the
influence of long primitives spanning the
whole bounding box of the input mesh
over the performances of distance and
intersection queries (accessible through
the benchmark menu).
• Loop subdivision. The surface is made
smoother through 1-to-4 triangle splits
and vertex relocations (not shown).
Benchmarks

Menu Benchmarks constructs an AABB Benchmark intersections

Benchmark intersections
Construct AABB tree...done (0.157 s)
tree with the mesh facets and allows Construct AABB tree...done (0.157 s)
Generates random queries...done (1.468 s)
Generates random queries...done (1.468 s)
benchmarking the number of intersection Benchmark do_intersect()
Benchmark do_intersect()
and distance queries per second with 198294 queries/s with segment
198294 queries/s with segment
197106 queries/s with ray
various random primitives. The user is 197106 queries/s with ray
217081 queries/s with line
217081 queries/s with line
invited to specify the duration of each test. 396296 queries/s with plane
396296 queries/s with plane
Benchmark any_intersected_primitive()
Benchmark any_intersected_primitive()
195489 queries/s with segment
195489 queries/s with segment
The sub-menu “against #triangles” 193271 queries/s with ray
193271 queries/s with ray
214415 queries/s with line
benchmarks various performance 214415 queries/s with line
386114 queries/s with plane
386114 queries/s with plane
indicators (memory occupancy, tree Benchmark any_intersection()
Benchmark any_intersection()
construction time, number of intersection 146888 queries/s with segment
146888 queries/s with segment
144476 queries/s with ray
and distance queries per second) against 144476 queries/s with ray
156757 queries/s with line
156757 queries/s with line
an increasing number of triangles of the 243704 queries/s with plane
243704 queries/s with plane
current mesh. The latter is obtained Benchmark number_of_intersected_primitives()
Benchmark number_of_intersected_primitives()
67802 queries/s with segment
through recursive longest bisection until 67802 queries/s with segment
57004 queries/s with ray
57004 queries/s with ray
the mesh complexity reaches 1M triangles. 55505 queries/s with line
55505 queries/s with line
All results are output in ASCII form to the 8676 queries/s with plane
8676 queries/s with plane
Benchmark all_intersected_primitives()
console window. Benchmark all_intersected_primitives()
65249 queries/s with segment
65249 queries/s with segment
53711 queries/s with ray
53711 queries/s with ray
52235 queries/s with line
52235 queries/s with line
6064 queries/s with plane
6064 queries/s with plane
Benchmark all_intersections()
Benchmark all_intersections()
47624 queries/s with segment
47624 queries/s with segment
38483 queries/s with ray
38483 queries/s with ray
35678 queries/s with line
35678 queries/s with line
2687 queries/s with plane
2687 queries/s with plane