Introduction

Point cloud data offers non-uniform coverage via points while a raster image displays complete coverage of an area using pixels. Converting a point cloud to a gridded raster results in easy to use data with a large amount of information. Using FME, you can convert a point cloud to a raster and customize your raster image by adjusting factors like pixel resolution or which point cloud component to display. The following exercises take you through different workflows step by step and explain how FME transformers can be used in different point cloud to raster translation scenarios.

1. Point Cloud To Grayscale Raster

Downloads

• pointcloud2grayscale.fmwt
• pc2gray-data.zip

This workflow will focus on using the point cloud component called intensity to create a raster output. Intensity is a quantitative indicator of the reflectivity of an object or how “bright” the return was during the point cloud data collection. For example, vegetation has high reflectivity and would return a high-intensity value while pavement or rooftops have low reflectivity and would return low-intensity values. If we visualize a point cloud using intensity only, we can create a grayscale raster.

1. Add the point cloud file with an LAS reader using default parameters.
2. Add the ImageRasterizer transformer and connect the LAS reader to the input port. The parameters should be changed so that:
   • Size Specification: CellSize
   • X Cell Spacing: 10
   • Y Cell Spacing: 10
   • Interpretation Type: Gray8
   • Background Color: 0,0,0
   • Anti-Aliasing: Yes
   • Input Component: Intensity
     

     

3. At this point, a grayscale raster can be created using a GeoTIFF writer with default parameters or more transformers can be added to the workflow to smooth the raster further as shown in the next steps.
4. Connect the raster from the ImageRasterizer to a RasterCellValueReplacer transformer and change the parameters to those seen in the image below where “Replace Values >=” and “Replace Values <=” are both set to 0, “New Value” is set to 127”, and “Replace Nodata” is set to Yes. This helps to remove some of the noise present in the raster image. The downside of this smoothing algorithm is degraded sharpness as we now have a lower contrast between the pixel values since we have replaced values 0 (black) with 127 (dark grey). The smoothing effect is less effective in areas with many bright pixels because they have values high above 127, so there is still an obvious contrast between the ‘holes’ and their surroundings.
   

   

5. Connect the previous output to a RasterResampler transformer and set the “Size Specification” parameter to “Percentage” with a Percentage of 101 and set “Interpolation Type” as “Bicubic”.
   

   

6. Copy the RasterResampler and paste it into the workspace. Change the set percentage of the second RasterResampler from 101 to 99.001 and connect it to the first RasterResampler transformer.
7. A smoothed raster output can now be written with a GeoTIFF writer using default parameters. Change the feature type definition to “Manual”. When the new window appears, change “Compression Method” to “Pack Bits” and “Photometric Interpretation” to “MinIsBlack”.

The result is a grayscale image. Below, from left to right are the source LAS file, rough output raster, and the smoothed output raster

An output image can often have some spots of NoData like where there is a body of water. In our case, we have spots of NoData where there is a lack of coverage due to the distance between points being larger than rasterization spacing. In the image below you can see where there are spots of NoData:

If there are many of these spots or if the output predominantly consists of NoData, consider using larger rasterization spacing like 20m instead of 10m (spacing is set in the ImageRasterizer). If there are few NoData pixels, try using smoothing algorithms such as resampling up and then down using the RasterResampler, as seen in steps 5-6 above. template workspace.

2. Point Cloud To DEM Raster

Downloads

• pointcloud2numeric.fmwt
• pc2numeric-data.zip

NumericRasterizer uses the Z coordinate to produce a DEM from a point cloud. The output will be effective so long as the assigned cell spacing is larger than possible irregular gaps in the point coverage. Due to the nature of point cloud data, these gaps are quite frequent and as a result, we should set a large cell spacing of 10m to avoid NoData holes in the DEM coverage. We cannot use smoothing as we did with intensity in the previous example because it will distort the surface making it appear as if it were hit with meteorites.

1. Add an LAS reader.
2. Simply attach the LAS reader to a NumericRasterizer transformer to create a raster output. The parameters required for the NumericRasterizer in our example are shown in the image below. Set “Size Specification” to “CellSize”, “X Cell Spacing” and “Y Cell Spacing” as 10 and “Tolerance” as 1. Try changing the cell spacing from 10 to 2 and compare the outputs. You will find that a spacing of 2 meters will result in many spaces:

   

3. A better option for rasterizing point clouds by elevation is the SurfaceModeller or RasterDEMGenerator. Both transformers can be used in the same way, however, the SurfaceModeller has additional abilities like generating contours or TINs. We have chosen to use the RasterDEMGenerator for this example. The RasterDEMGenerator makes the same DEM as the NumericRasterizer, however, it has more intelligence under the hood and is able to interpolate holes giving much smoother output even with small spacing. The parameters required for the RasterDEMGenerator in our example are shown in the image below. “Surface Tolerance” is set to 0.0, and “Output DEM X Cell Spacing” and “Output DEM Y Cell Spacing” are both set to 2.

   

4. Run the workspace with Feature Caching and inspect the output of the NumericRasterizer and the RasterDEMGenerator.

The images below show the original point cloud, the raster produced by the NumericRasterizer (10 meter spacing), and the raster produced by the RasterDEMGenerator respectively:

Data Attribution

• The data from the first exercise originates from data made available by the Ohio Geographically Referenced Information Program. The GIS Support Center maintains enterprise and site licenses for commercial data sets that are supportive of the Ohio Enterprise.
• The data from the second exercise originates from data made available by West Virginia View. They are sponsored by AmericaView and the USGS.