Answer
With FME 2011, we added a new type of geometry called
Point Cloud.
Wikipedia defines a point cloud as a set of vertices in a three-dimensional coordinate system.
We can think about a point cloud as a collection of multiple points, however, that would be oversimplifying things. Surprisingly, when lots of points are brought together into a single feature, they start to show some new qualities.
For example, point cloud is obviously a vector based structure - each point has its XYZ coordinates, and some attributes called
components in FME. Components can represent time, flight line, intensity (how much light returns back from a point), color, etc. However, in many respects, point clouds behave like rasters. As Dale Lutz once said, "
point cloud is a badly behaved raster". We can apply many typically raster transformations to point clouds. To a certain extent, we planned our development based on our raster capabilities, and our raster programmers were greatly contributing to making FME point cloud friendly.
So, for example, we can thin (or decimate) a point cloud, that is, reduce the number of points in it. We can compare this operation with raster resampling. Combining point clouds together is similar to raster mosaicking, and making point clouds from other geometries can be compared with rasterization.
In GIS, point cloud geometry is often known, rather mistakenly, as
LiDAR. LiDAR is a technology for making point clouds, and from this point of view, LiDAR data, that is, data produced with LiDAR technology, is a valid term, however, as you will see later, point clouds can be made from just about any other geometry. This is why we call our new geometry "point cloud", and not "LiDAR", and our new transformers dealing with it have "PointCloud" prefix (
PointCloudCombiner,
PointCloudSplitter).