Point clouds may contain color information about each point, and the modern LiDAR sensors allow collecting colors, however, most of the point cloud datasets available today don't have this information.


One of our customers gave us a wonderful idea - why don't we try to color point clouds with widely available orthophotos.


I tried it in FME2011. I found, this is possible, however, with some limitations. Let's have a look what FME is doing by default when we supply both a point cloud and an ortho image to the PointCloudCombiner transformer that combines all kinds of geometries into point clouds:


Source: LAS and GeoTIFF    


Result: Single LAS


On the left side we see a point cloud colored only in DI (it does not have color components) and an image below. The output consists of single point cloud - the bottom flat part at elevation 0 is made of pixels of the ortho, and the top is the original point cloud. Note that DI in FME 2011 cannot color point clouds that have R, G, and B components, this is why the top is black. I don't think, there is much practical value in this conversion.

If we could only transfer colors from the ortho, dropping the points, we would get something more interesting. There is no direct way of doing this in FME 2011. The trick is to rasterize the point cloud, make a special RGBDEM raster, and turn it into a point cloud, that is, to follow the steps of our scenario 15:



Note that there is the custom PointCloudZClipper before the writer - Point Cloud rasterization often produces lots of nodata pixels (see Point Cloud Scenario 11. Point Cloud Rasterization) - these pixels will have some valid RGB value and nodata for the DEM band. Thus, these pixels will not be elevated to the original point cloud body:



We can get rid of them by clipping the point cloud to the valid range, which will give us the desired result:



For FME 2012, there is a plan to do a better job with coloring point clouds using ortho images, so that the points can be colored directly, without rasterization step. However, the final decision about time frame and details of implementation are still being discussed.