Many of the parameters used below rely on the average point distance between individual points in a point cloud. The average point distance is the mean distance between every point in a cloud and its nearest neighbor in a Euclidean sense.
Another common measure used is the average neighborhood size. This integer value is the mean of the nearest neighbors of each point in the point cloud when a certain search radius is used to gather neighbors. The search radius used is the average point distance referred to earlier.
The following summarizes the meshing steps and their parameters. Generally the default values for the parameters should produce good results, but some experimentation with parameters may be required for certain projects. You can use the ‘Set Defaults’ button at the top of the dialog to reset any adjustments made.
• Register and Merge - Register and merge two or more Point Clouds. Improve line-up, reduce data in the overlap areas and then produce one combined Point Cloud or Mesh.
• Merge Only – In certain cases (e.g. point clouds with noise), the registration step can result in mismatches and a distorted point cloud. If this is checked the registration step can be skipped and point cloud data is only merged.
• Include selected object points – if a set of points (regular or SmartPoints) are selected while the Cloud/Mesh Options dialog comes up, the selected points will be included in the Meshing steps. This can be useful when, for example, you want to just triangulate a set of SmartPoints, without necessarily generating a DSM point cloud first.
• Point Normal Estimation – Triangulation relies on estimated ‘point normals’ (the point normal is a line or vector that is perpendicular to the surface being defined) in order to help accurately define a surface. The closer the point normals are to the eventual surface, the better the triangulation. The method used to define the normal can be set here, where you can choose the plane (XY, YZ, or XZ), or the view direction (vector from point to camera). In most cases where the cameras point directly at the surface (e.g. aerial photos), the XY plane is usually most appropriate, but other options are available for other types of triangulations. The ‘View direction’ setting may be most appropriate for models where the cameras wrap around the object.
• Clear redundancy – Remove duplicate points that are close to each other within the mesh.
• Minimum distance – Points within this distance of each other are considered redundant and will be removed. The average distance between points in the mesh multiplied by this value.
• Filter Isolated clusters – Removes isolated points and small point clusters from the mesh that are disjointed from larger clusters of points. Such isolated points and clusters are either noise or represent surfaces with inadequate point density.
• Cluster Size Threshold – Clusters with a point count less than this threshold will be removed. Enter the percentage value used to apply to the total number of points in the mesh. E.g., a value of 5 will result in a cluster size threshold of 5% of total point count.
• Cluster Size Distance – Minimum radius (in units of average point distance) that defines an isolated cluster. The average distance between points in the mesh multiplied by this value. E.g., a value of 5 will result in a minimum radius distance of 5 x average distance.
• Additional Filters – A number of additional filters to reduce noise and smooth the point cloud. These filters can be used to make the point cloud more suitable for triangulation.
• Statistical Outlier Removal – Removes points that do not agree with the statistical spread of the points in the point cloud. A statistical distribution is built by calculating the average point distance between the neighbors of each point. This distribution is assumed to be Gaussian and is parameterised using the mean and the standard deviation. Once this distribution is built, all points whose average distance from points in its neighborhood differ from the mean by more than some multiple of the standard deviation are eliminated.
• Neighborhood Size - The size of the neighborhood over which to build a statistical distribution (in units of average neighborhood size). This value multiplied by average neighborhood size.
• Distance Threshold - The distance threshold over which the point is eliminated (in units of standard deviation of a Gaussian distribution). A value of 1 implies that approximately 67% of all points will be within this distance.
• Neighbor-based Outlier Removal – Removes points which do not have a sufficient number of neighbors within a specified search radius.
• Neighborhood Size - The size of the neighborhood (in units of average neighborhood size) which is the threshold for a point to be retained.
• Search Radius - The search radius over which to look for neighbors (in units of average point distance).
• Smooth Points: Locally fits an n-th order polynomial around point neighborhoods and fits a smooth surface to the point cloud. This allows a subsequent resampling of the point cloud to obtain smoothed points.
• Smoothing Amount - The amount of smoothing that will be performed.
• Iterations - The number of iterations to be performed for smoothing.
• Point Decimation – Reduce the number of points while retaining the shape of the point cloud. A voxel grid is built over the bounding cube of the point cloud. In every voxel, all the points are replaced by their centroid. This preserves the shape of the point cloud but point positions are not preserved. An alternative version allows preservation of the point positions.
• Grid Size - The size of an individual grid (in units of average point distance).
• Preserve Point Positions - Preserves the positions of the points.
• Triangulation – Triangulate a point cloud to form a surface.
• Trim Mesh by bounds – Ensures the triangulated surface lies within the bounds of the original point cloud. Use when the smooth extension of the mesh is not desired.
• The Trim threshold allows you to control the aggressiveness of the trimming. Increasing this value will remove more triangles during the trimming stage.
• Mesh sharpness - The amount of detail/sharpness of the triangulation. A larger number will create more triangles giving sharper details. It can be used to trade-off quality for mesh size (in terms of number of triangles).
• Decimate Triangles – Reduce the number of triangles while retaining the overall shape.
• Percent – The percentage of triangles to be removed.
• Additional triangle filters – More operations to refine the triangulated mesh.
• Smooth triangles – Adjust the triangles in the mesh to remove small bumps and imperfections.
• Number of iterations – The number of smoothing iterations. The larger the value the smoother the result.
• Fill holes – Fill enclosed areas that were not triangulated due to missing point data.
• Maximum hole size – Maximum size of hole to fill. The average distance between vertices in the mesh is multiplied by this value.
• Fill largest hole - Uncheck for 2.5D surfaces as this fills the largest hole (assumed to be an open boundary) in the mesh.
• Contours - Creates contour lines at regular intervals on the surface.
• Major Spacing - the distance between major contours in the direction of the axis. This value is in project units. Thickness and style of major contours can be set on Preferences – Aerial/UAV.
• Minor Spacing - the distance between minor contours (between major contours) in the direction of the axis. This value is in project units. Usually best to space minor contours as even divisions of major (e.g. major 100, minor 10) but any spacing is acceptable. Thickness and style of minor contours can be set on Preferences – Aerial/UAV.
• Axis - which of the three coordinate axes to step along. Note that a Project Rotation (see Importing and Setting Up Coordinate Systems) should be set up before running this step to get best results. For example if you wanted height contours (as we are used to seeing on topographic maps) you would set up a project rotation such that the X,Y plane was the ground plane and Z was ‘up’. Pick the ‘Z axis’ option here and the height spacing required in the Spacing parameter above.