PointCloudSimplifier

Point cloud features.

Point cloud features with a simplified subset of the original points.

Point cloud features containing all points not included in the simplified point cloud.

Non-point cloud geometries are output via the <Rejected> port.

Medial Axis Transform: Perform simplification based on the Medial Axis Transform (MAT). This method estimates normals for each point and uses a ball-shrinking algorithm to approximate the MAT. Then it samples the point cloud based on each point’s Local Feature Size (LFS), defined by the shortest distance from the point to the medial axis. Points in areas of high curvature (low LFS) are sampled at a higher density than redundant low curvature (high LFS) points.

Specifies the desired level of simplification, relates LFS values to sampling density. Higher values will result in lower overall density. Typical values range from 0.01 to 0.8.

Specifies the minimum sampling density for grid cells (in points per square unit). A value of 0 means that the sampling density is not bounded from below. This can be useful to maintain a minimum level of detail in nearly flat regions which would otherwise be removed entirely because of plane detection.

Specifies the maximum sampling density for grid cells (in points per square unit). A value of 0 means that sampling density is not bounded from above.

Linear: Use the unmodified LFS values for each point during simplification.

Quadratic (LFS^2): Use the square of the LFS value for each point during simplification. This will increase the effect that curvature has on point sampling.

Specifies the cell size for grid-based LFS simplification. During simplification, the point cloud is divided into a uniform grid of this size. The sampling density is calculated for each cell based on the average LFS values of its contained points. This should be as small as possible, but large enough to have an average of 5-10 points per cell. Selecting Auto will let FME estimate a reasonable cell size based on the bounding box and number of points in the input point cloud.

Specifies the initial radius used by the ball-shrinking algorithm. Larger values will allow for the detection of larger shapes when approximating the MAT. Selecting Auto will let FME estimate a reasonable initial radius based on the bounding box and number of points in the input point cloud.

Specifies the number of nearby points to consider when estimating point normals. Use higher values with denser inputs to produce more consistent normals. Typical values range from 5 to 20.

Specifies the minimum angle (in degrees) between the defining points for a medial ball during a ball-shrinking iteration. Iteration stops when the angle would fall below this threshold. A value of 0 means that no ball preservation takes place. Typical values range from 0 to 30.

Specifies the minimum angle (in degrees) between the defining points for the initial medial ball. If the angle on the first iteration is below this threshold then no medial ball is assigned. A value of 0 means that no plane detection takes place. Typical values range from 0 to 40.

Specifies the number of nearby points to check against the Bisector Threshold. Higher values will result in fewer noisy MAT points but will also limit the ability to detect small features. Typical values range from 1 to 5.

The Bisector Threshold is used to clean the MAT points before LFS computation. Lower values mean more aggressive cleaning which means more robustness to noise in the MAT, but fewer features will be detected. Typical values range from 0.1 to 10.

Use Interior MAT Only: Only consider the interior MAT when computing LFS values.

Use Interior and Exterior MAT: Consider both the interior and exterior MAT when computing LFS values.