Two workflows are defined to automatically generate a single seamless surface from multiple disparate and potentially overlapping point clouds. The input data may contain inconsistencies as the datasets are obtained at different points in time, with different acquisition methods having different properties. The most extensive workflow offers an automatic deconfliction treatment, a process that until now has required a manual selection of data sets. To apply this workflow, the data sets must be prioritized according to trust. The deconfliction service detects inconsistencies and produces a cleaned point cloud where the initial point cloud of highest priority is kept, along with sub point clouds from other data sets that are found to be consistent with the high priority point cloud. If the data sets don't contain priority information, deconfliction cannot be performed, and the simpler workflow is applied to create the surface.

Two workflows are defined to automatically generate a single seamless surface from multiple disparate and potentially overlapping point clouds. The input data may contain inconsistencies as the datasets are obtained at different points in time, with different acquisition methods having different properties. The most extensive workflow offers an automatic deconfliction treatment, a process that until now has required a manual selection of data sets. To apply this workflow, the data sets must be prioritized according to trust. The deconfliction service detects inconsistencies and produces a cleaned point cloud where the initial point cloud of highest priority is kept, along with sub point clouds from other data sets that are found to be consistent with the high priority point cloud. If the data sets don't contain priority information, deconfliction cannot be performed, and the simpler workflow is applied to create the surface.

To handle large data volumes, and use distributed computing, tiling of input data is performed. The tiles are distributed in the cloud and an LR B-spline surface is generated for each tile. LR B-spline is a smooth surface type that is particularly well suited when the input point clouds represent a smooth terrain (topography or bathymetry). This approach most often lead to a considerable data reduction (typically two orders of magnitude) compared to the input data. In a final step the LR B-spline surfaces of each tile are stitched to a global surface. The generated surface can be sampled to generate grid based surfaces representations such as GeoTIFF where the number of points generated is controlled by user specified parameters.

The Figure to the left shows an example of a surface generated by the workflow where no deconfliction is applied. To avoid modelling noise, a restricted number of iterations in the adaptive surface approximation algorithm applied by the workflow is performed. The final surface consists of 15 individual LR B-spline surfaces glued with C1 continuity. The total surface size is about 1% of the size of the input data.

The first Figure below shows the result of the workflow where deconfliction is included on a small survey only containing 10 individual blocks with a total input data size of 28 Mbyte. The second Figure below  shows details of the resulting surface, and points sorted by the deconfliction algorithm, i.e. remaining and removed points. The images are generated by the result IQmulusViz.

 

RESULTING SURFACE SET AND DATA POINTS WITH DISTANCE INFORMATION, ALL POINTS (LEFT), CLEANED POINT SET (MIDDLE) AND REMOVED POINTS (RIGHT)
 

A DETAIL PERSPECTIVE WITH SEVERAL OVERLAPPING DATA SURVEYS, ALL POINTS (LEFT), KEPT POINTS (MIDDLE) AND REMOVED POINTS (RIGHT).