Visualizing Data Layers


TileDriver Visualize is all about viewing geospatial data and imagery. These data are presented as layers in the Visualize sidebar.


Each layer is listed with its name, a thumbnail representing its display type, and a switch. Clicking the switch displays the layer on the map. Clicking on the thumbnail image displays the layer and zooms you to the layer on the map.

Once the switch is in the "on" position, you will see other options. The airplane icon will zoom the map to the area where the layer exists. The gear icon opens a dropdown menu where you can set layer options (see below) as well as download the file associated with the layer (if any).

NOTE: TileDriver Basic accounts do not have a download file option.

Finally, the slider allows you to set the opacity of the layer.

Setting layer options

A few layer types have options you can change in order to alter the way the layer displays:

  • Vector layers with the Hexbin display type
  • Vector layers with the Heatmap display type
  • Raster layers

Hexbin layer display

When you activate a Hexbin Vector layer, you will see something like the following:


The purpose of binning data is to analyze relative concentrations of something, based on the quantity of measurements or some other data value. This particular example is a vector representation of a neural network result, showing areas of Washington, DC that contain high or low concentration of forest based on the network's scoring of overhead imagery.

As you can see, the hex cells are colored with different shades of green representing higher levels of forest (darker green) to lower levels of forest (lighter green). This is great, but what if you wanted to view the data differently? Maybe the hex cells are the wrong size for the data, or the color scheme is non-intuitive.

To change the display of the hexbins, click the gear icon and choose "Set layer options" from the menu. You will see a screen like this one:


TileDriver gives you lots of options to modify the display of Hexbin layers. The first thing you'll notice is that the radius of the hex cells can be changed. Right now it's set to 400 meters. Well, meters makes sense because the neural network was scoring image chips of a specific size. However, 400 meters clearly looks too big. Let's change it to 200 meters.


This looks a lot better already. The opacity has been lowered in this screenshot so you can see the underlying satellite image. The hex cells conform much more closely with the areas of forest (primarily Rock Creek Park, as the locals will know right away). We kept the radius type as meters, but you can also change it to pixels if the data warrant it. A pixel-based radius will cause the hex cells to remain a fixed size as you zoom in and out.

The next thing you'll see is a histogram of the data values corresponding to how high an image chip was scored for forest. A slider bar displays the low and high values, and the color scale is also displayed below the histogram.

You probably notice that the handle on the slider bar on the high end of the range is pretty far to the right of where the data values trail off. It might be better to slide that to the left so we see a better representation of the data. Right now the high end is set to 0.18; let's slide it over to 0.14. This is the result:


If you compare this to the previous screenshot, you'll see that some of the areas of forest are deeper green indicating a higher score from the neural network. This seems to make sense; you can turn the layer on and off to validate.

Next you'll notice that you can change the color scheme used to display the hex cells. Here we're using "Sequential Greens" with 9 shades, and that seems to do the trick. TileDriver comes with full support for ColorBrewer color schemes; just start typing the name of one of the schemes and you'll see sample palettes.

Finally, you'll notice that the hex cells can be styled in different ways. In our case, maybe turning off the stroke (the line around the hex cells) would look better. If we turn that off, we see this:


I believe our work here is done.

Heatmap layer display

Heatmaps are similar to Hexbins in that they display relative concentrations of data. Here's an example of one, using the same neural network as the Hexbin example above, but this time looking for areas of "medium residential" development:


Looking at this image from a high altitude, it appears to be a little too inclusive; meaning, that the heatmap rates too much of the image as having a high score for residential. If we zoom in to the West side of the image, you can see this more clearly:


Parts of Rock Creek Park that sit between residential areas are showing as residential (the red color of the heatmap indicates a high score for residential, contrasted with blue for low scores). We can see this in particular in the three highlighted areas shown below:


Thankfully TileDriver gives us ways to improve this display, just like with Hexbin Vector layers. Here is the layer options screen for Heatmaps:


We can see that the radius for generating the heatmap is 800 meters. That probably explains what we're seeing. What if we dialed that down to 600?


This is a much better result. The heatmap contours match the line of residential development more closely. We could continue to play with this value until it represents the neural network results as well as possible.

The "blur radius" setting determines how sharp the line is from one color to the next in the heatmap color range. Right now it's set to 4; let's dial that up to 10:


You can see how this alters the resulting display. Experiment with this value to see how it brings out more information in your data layers.

Raster layer display

TileDriver gives you a lot of control when displaying raster image layers in Visualize. This capability really shines when using multispectral images (MSIs; if you don't know what those are, check out this Wikipedia article).

Let's use the following WorldView-2 satellite image to demonstrate what you can do with TileDriver Visualize:


This probably looks pretty useless to you right now. However, there's a lot of data within this murky picture. We just need to compensate for a few things to bring it out. Let's use the layer options screen (found under the gear icon):


Let's start with a feature that should look familiar from the Hexbin Vector layer options: the histogram of color values. With raster images, we see histograms for red, green and blue values.

As this is a satellite image, you have to assume a fair amount of atmospheric interference. This accounts for the large amount of "noise" in these histograms (the "signal" is represented by the peaks of red, green and blue values to the left of the graphs; the "noise" is everything outside of that).

Like we did for the Hexbin layer earlier, we can slide the high and low values of the color scale for each of the three bands toward the signal:


Now we can see a much clearer image:


However, this is an 8-band image with lots of wavelengths of light being captured by the sensor1. What if we want to use some of the other bands to tease out additional insights from this image?

To use other bands, we just need to change the values in the "red band," "green band" and "blue band" fields to the bands we want to look at. Let's say we want to use the Red Edge band (band 5; this is in the infrared area of the spectrum, very close to the edge of the visible spectrum) and use the normal green band (band 2), and normal blue (band 1).

Type the new band values into the form and click "Save Changes". By changing those values, we get the following:


Now the image is murky again, but we know how to clean that up. Just adjust the RGB band sliders closer to the signal (notice how the histogram shapes have changed; remember that you're looking at different bands of image data than before):


And then we get the following:


This is an image we can really use for analysis. For example, we can clearly see that forested areas reflect differently in the Red Edge band than the normal red band, as compared with non-forested areas.

Congratulations! You're now an imaging scientist.

  1. If you want to learn more about the WorldView-2 sensor, including what bands are captured, see: