us-iale2017.html

<section>
    <h2>3D rasters are powerful</h2>
    <h3 style="margin-top: -0.5em">and you already know how to use them</h3>
    <h3 style="margin-top: 0.5em">
        Vaclav (Vashek) Petras,</h3>
    <h4 style="margin-top: -0.7em">Anna Petrasova and Helena Mitasova
    <p class="title-foot">
        <a href="http://www.ncsu.edu/" title="North Carolina State University">NCSU</a>
        <a href="http://geospatial.ncsu.edu/osgeorel/" title="NCSU GeoForAll Lab">GeoForAll Lab</a>
        at the
        <a href="http://geospatial.ncsu.edu/" title="Center for Geospatial Analytics">Center for Geospatial Analytics</a>
        <br>
        North Carolina State University
    </p>
    <p>
        US-IALE, Baltimore<br>
        April 9-13, 2017
    </p>
</section>


<section>
<h2>3D rasters</h2>
<p>
Usage:
soil properties,
concentrations in water bodies,
properties of air mass,
hyperspectral data,
analysis of spatio-temporal data,
forest and habitat properties,
exploration of simulation scenario space,
etc.
</p>
<ul>
    <li><strong>3D raster</strong> - Voxel space - 3D voxel grid - Voxel model
    <li><strong>[3D] cell</strong> - Voxel - 3D voxel unit - Volumetric pixel
</ul>
<img src="img/raster3d_layout.png" class="stretch">
<p class="credit">
    Image credit:
    <a href="https://grass.osgeo.org/grass72/manuals/raster3dintro.html">GRASS Development Team</a>
</p>
</section>


<section>
<div class="left">
    <h3>Series of 2D rasters</h3>
<ul>
    <li>2D does not explicitly represent and model 3D relationships
    <li>2D rasters accessible individually
</ul>
<img src="img/raster2d_series.png" style="width: 100%;">
</div>
<div class="right">
    <h3>3D raster</h3>
    <ul>
        <li>explicit 3D relationships
        <li>unified access to all 3 dimensions
        <li>needs to be sliced to get 2D rasters
    </ul>
<img src="img/raster3d_stack.png" style="width: 100%;">
</div>
</section>


<section>
<h3>Third dimension</h3>
<ul>
    <li>The first two dimensions are the same as in 2D.
    <li>Third dimension can be in space, time, or scenario space.
    <li>Mind the possible anisotropy in the Z direction.
        <br><small>(different from 2D rasters where all directions are isotropic most of the time)</small>
</ul>
<img src="img/axes.png" class="stretch">
</section>


<section>
<h2>Working with 3D rasters</h2>
Not that different from 2D rasters, images, and matrices
</section>


<section>
<h2>Python</h2>
<ul>
    <li>using NumPy array
</ul>
<pre class="python" style="width: 50%;"><code>x = array([[[ 0,  1,  2],
            [ 3,  4,  5],
            [ 6,  7,  8]],
           [[ 9, 10, 11],
            [12, 13, 14],
            [15, 16, 17]],
           [[18, 19, 20],
            [21, 22, 23],
            [24, 25, 26]]])
x.sum(axis=0)
# array([[27, 30, 33],
#        [36, 39, 42],
#        [45, 48, 51]])
</code></pre>
<img src="img/logos/python.svg" class="stretch">
</section>


<section>
<h2>R</h2>
<ul>
    <li>using array
</ul>
<pre class="r" style="width: 50%;"><code>x = array(1:40, dim=c(2,4,5))
# , , 1
#
#      [,1] [,2] [,3] [,4]
# [1,]    1    3    5    7
# [2,]    2    4    6    8
#
# , , 2
# ...
x[2,3,3]
# 22
x[2,3,3] = 0
</code></pre>
<img src="img/logos/r.svg" class="stretch">
</section>


<section>
<h2>Octave</h2>
<ul>
    <li>using 3D matrix/array
</ul>
<pre class="octave" style="width: 50%;"><code>x = reshape(1:8, 2, 2, 2)
a(2,1,1) = 5
b = 3 * a
# b =
#
# ans(:,:,1) =
#
#     3    9
#    15   12
</code></pre>
<img src="img/logos/octave.svg" class="stretch">
</section>


<section>
<h2>GRASS GIS</h2>
<p>
    <em>general-purpose geospatial analysis and remote sensing tool</em>
</p>
<ul>
    <li>
        data types:
        <ul>
            <li>vectors
            <li>2D rasters
            <li><strong>3D rasters</strong>
        </ul>
    <li>
        container types:
        <ul>
            <li>imagery groups
                <ul>
                    <li><strong>series of 2D rasters</strong>
                </ul>
            <li>spatio-temporal datasets
                <ul>
                    <li><strong>series of 2D rasters</strong>
                    <small>(or vectors or 3D rasters)</small>
                </ul>
        </ul>
</ul>
<br>
<img src="img/logos/grassgis.svg" class="stretch">
</section>


<section>
<h2>Moving window</h2>

<h3>2D</h3>

<pre class="bash" style="width: 100%;"><code>r.neighbors input=a output=b method=average size=5
r.colors map=b color=viridis
</code></pre>

<h3>3D</h3>

<pre class="bash" style="width: 100%;"><code>r3.neighbors input=a output=b method=average window=5,5,5
r3.colors map=b color=viridis
</code></pre>

</section>


<section>
<h2>3D raster algebra</h2>

Addition

<pre style="width: 55%;"><code>b = a + 2
</code></pre>

Conditional expression

<pre style="width: 55%;"><code>b = if(a > 20, 1, null())
</code></pre>

Accessing neighboring cells

<pre style="width: 55%;"><code>b = (a[0,0,-1] + a[0,0,1]) / 2
</code></pre>

</section>


<section>
<h2>Visual programming</h2>
<ul>
    <li>graphical modeler
    <li>graphical raster calculator
</ul>
<br>
<img src="img/graphical_modeler.png" class="stretch">
</section>


<section>
<h2>Example: Terrain evolution</h2>
<img src="img/jr_animation_2d.gif" class="stretch">
</section>


<section>
<h2>Example: Terrain evolution</h2>
<img src="img/jr16animation.gif" class="left">
<img src="img/jr20animation.gif" class="right">
<img src="img/raster3d_from_2d.png" class="stretch">
<p class="credit">
    <small>
    Image credit:
    <a href="https://grass.osgeo.org/grass72/manuals/r.to.rast3.html">GRASS Development Team</a>
    </small>
</p>
</section>


<section>
<h2>Example: Behavioral patterns</h2>
presence of people over time in camera view
<img src="img/amos_animation_2d.gif" class="stretch">
<p>
Petrasova A., Hipp, A.J., Mitasova, H.,
<em>Using space-time cube for visualization and analysis of active
transportation patterns derived from public webcams</em>.
ICC 2017, Washington DC 2-7th July
</p>
</section>


<section>
<h2>Example: Behavioral patterns</h2>
spatio-temporal data representing movement of people in urban environment
<br>
<img src="img/amos_animation_3d.gif" class="stretch">
<br>
Petrasova A., Hipp, A.J., Mitasova, H.,
<em>Using space-time cube for visualization and analysis of active
transportation patterns derived from public webcams</em>.
ICC 2017, Washington DC 2-7th July
</section>


<section>
<h2>Example: Vegetation structure</h2>
3D vegetation structure derived from lidar point clouds
<small class="glow">
    Petras, V., D. J. Newcomb, and H. Mitasova.
    <em>Generalized 3D fragmentation index derived from lidar point clouds.</em>
    In: Open Geospatial Data, Software and Standards [in print]
</small>
<img src="img/fragmentation_paper.png" class="stretch box-glow">
<p class="credit">3D version of forest fragmentation index by Riitters et al.</p>
</section>


<section>
    <h3>3D index of 3D raster</h3>
    <img class="stretch" src="img/raster3d.png">
    <p class="credit">Petras et al. 2017 [in print]</p>
</section>


<section>
    <h3>Point cloud</h3>
    <img class="stretch" src="img/points3d.png">
    <p class="credit">Petras et al. 2017 [in print]</p>
</section>


<section>
    <h3>Binning, moving window &amp; raster algebra</h3>

<pre class="bash" style="width: 100%; font-size: 75% !important;"><code># Bash
r3.in.lidar input=points.las n=count base_raster=ground -d class_filter=3,4,5
r3.neighbors input=count output=average method=average window=3,3,3
r3.mapcalc expression="buffer = if(average > 0, 1, 0)"
</code></pre>

    <img class="stretch" src="img/profile2d_presence.png">
    <img style="width: 10%;" src="img/count_legend.png">
    <p class="credit">Petras et al. 2017 [in print]</p>
</section>


<section>
    <h3>Fragmentation index</h3>

<pre class="python" style="width: 97%; font-size: 85% !important;"><code># Python
import grass.script as gs
gs.run_command("g.extension", extension="r3.forestfrag")
gs.run_command("r3.forestfrag", input="buffer", output="fragmentation")
</code></pre>

    <!-- stretch does not work well here -->
    <img style="width: 100%; margin-bottom: 0em;" src="img/profile2d_ff.png">
    <br>
    <img style="width: 20%; margin-top: 0em;" src="img/ff_legend.png">
    <p class="credit">Petras et al. 2017 [in print]</p>
</section>


<section>
    <h3>Profile of 3D raster</h3>
    <img class="stretch" src="img/profiles3d.png">
    <p class="credit">Petras et al. 2017 [in print]</p>
</section>


<section>
    <h3>Back to 2D</h3>

<img style="width: 20%;" src="img/count_categories.png">

<pre class="bash" style="width: 75%; font-size: 80% !important;"><code># Bash
g.extension r3.count.categories
r3.count.categories input=fragmentation output=count
g.list type=raster pattern="count_*"
</code></pre>

    <img class="stretch" src="img/relative_count.png">
    <p class="credit">Petras et al. 2017 [in print]</p>
</section>

<section>
    <h3>Conclusion</h3>
<ul>
    <li>there is a third dimension
    <li>3D rasters can handle it
    <li>use the tool you know, or pick one
</ul>

<p style="text-align: left; width: 40%; margin-left: auto; margin-right: auto;">
    <img src="img/logos/grassgis.svg" style="width: 10%; vertical-align: middle;">
    GRASS GIS
    <a href="http://grass.osgeo.org">grass.osgeo.org</a>
    <br>
    <img src="img/logos/python.svg" style="width: 10%; vertical-align: middle;">
    Python
    <a href="http://python.org">python.org</a>
    <br>
    <img src="img/logos/r.svg" style="width: 10%; vertical-align: middle;">
    R
    <a href="https://www.r-project.org/">r-project.org</a>
    <br>
    <img src="img/logos/octave.svg" style="width: 10%; vertical-align: middle;">
    Octave
    <a href="https://www.gnu.org/software/octave/">octave.org</a>
</p>
<p>
    <a href="https://github.com/wenzeslaus">github.com/wenzeslaus</a>
</section>