Directions Magazine (DM): Let’s talk about hybrid modeling of point clouds, raster imagery and vector geometry. Most software solutions can handle this integration already. What makes Bentley’s software unique and what’s the real advantage of the hypermodel that somehow makes the navigation intuitive?
Benoit Fredericque (BF): So far, the tools able to combine point cloud and vector data were mainly dedicated tools focusing on one paradigm: building geometric models from point clouds.
Bentley's vision is to enable point cloud usage to people who are not point cloud experts in paradigms where point clouds are used by the industry experts, as is. For example, in a traditional approach, one scans a plant area, creates a geometric model, imports the geometric model in the plant design application and performs the new design. The designer assesses design feasibility using the imported geometric model that must be exhaustive and is time-consuming to create.
The direct use of point cloud data in the design application reduces the amount of digitizing either by using directly the point cloud as is (the point cloud is directly analyzed to assess the design feasibility) or by extracting geometries only for specific areas requiring it (driven by designer knowing and discovering his needs).
Bentley's unique value proposition is to integrate point cloud data natively in the design applications (MicroStation, Bentley Substation, OpenPlant, InRoad). This integration enables new workflows, as well as taking advantage of other innovations such as hypermodels.
The concept of hypermodels combines 3D and 2D representations and links these two medias dynamically. The full 3D models provide great context and representation and are easier to understand than the 2D representations, focusing on specific areas. Hypermodels enable the smooth integration between those representations and apply to point cloud.
DM: Tell us more about scalable terrain models. Is Bentley disaggregating point clouds first? If users want to employ a higher spatial resolution model, they then use more points? Is there some compression technology involved in this?
BF: Scalable terrain model (STM) generates a multi-resolution spatial index of all 3D points and breaklines referenced. This index is then used to compute a surface on-the-fly at the right density\level of detail based on the user viewpoint. There is no decimation of the data input when the user is zoomed enough but we compute a decimated representation as soon as the user zooms out.
STM references data from different data sources including point cloud files. This avoids duplicating data. Each time the data source is changed, we detect it and ask the user if the index must be updated.
DM: Explain more about the challenge of integrating terrestrial and airborne-acquired LiDAR data. Is this something that Pointools brings to Bentley’s ability to use LiDAR data?
BF: There are several challenges in integrating aerial LiDAR and terrestrial depending on what part of the workflow you look at (such as registration of the two together in a consistent way).
Bentley’s support of point cloud starts after the registration and the main challenge is to be able to efficiently handle point cloud in a way that is agnostic of the type of sensor used.
For example, we frequently see data management strategies that apply only to outdoor data for medium and large coverage (ex: tiling files). Such solutions can work with aerial data but do not deal with the complexity of a subway station.
The vortex engine used in Bentley’s design application to display point clouds, as well as ProjectWise point cloud streaming service, is agnostic of the device used to acquire the data. These two key capabilities have been enabled by the Pointools team.
DM: Tell us about the ProjectWise point cloud services. Is the streaming of points on-demand a local or cloud service? Is there a vision to support point cloud technology as a cloud service?
Faraz Ravi (FR): The point cloud streaming works by streaming incremental view-based sets of points to a user’s machine. Typically the server (which hosts the point cloud files) is on an enterprise's internal network, which may span multiple sites or even countries. The integrated approach of managing point cloud data alongside design models and other assets within ProjectWise with full rights management and other project/file management capabilities is one of the key benefits of this approach for point cloud streaming and sharing. So the point cloud services are not as a "cloud" based service.
DM: Finally, let’s talk about point cloud classification editing. Are point clouds being classified on-the-fly as a particular object or feature? Is the snapping of vectors to point clouds a means to provide more spatial precision to a 3D model?
FR: Points within a point cloud can have a number of attributes in addition to the mandatory x, y and z values. These are usually color r, g and b, intensity and classification. The point's classification value is often used to indicate that a point is a ground, foliage, road, building, etc. point (for airborne LiDAR workflows) - similar to layers. Alternatively, classification can be used, for example, to indicate points belonging to objects in a plant environment. Descartes enables the editing and display of classification values via a number of advanced tools.