Geometry has its place again! This time in the form of segmented point clouds.

The journey the spatial data industry has traveled over the past 20 years has been somewhat remarkable. We started with simple CAD primitives, points, lines, and planes. We then progressed to volumes to provide 3D context. These volumes became “intelligent” with an ability to reference data directly on the volumetric object itself and the concept of BIM was born.?The industry then sat back in a holding pattern for 10 years whilst the hardware and cloud technology progressed to a point that it is at today. Recently, the concept of the geospatial digital twin was born.?There has been one constant through the 20-year spatial data journey, geometry.

Geometry has been the foundation of all spatial information ever since CAD was invented in 1963 by Ivan Sutherland and Sketchpad. And geometry will inevitably be the foundation of spatial information for some time to come. However, the format of geometry will be forever changing based on the technology and purpose of spatial data. As we move to an object orientated point cloud data objective, we need to appreciate what this means and how this will enable brownfields and Greenfields assets expedite the digitization process. For many years, there has been a misconception that the only solution to digitizing a brownfields asset is to back model from point cloud data. However, this is now starting to change. The introduction of simplified cheaper alternative mobile mapping and scanning technologies, plus the emerging technology of object segmentation capabilities to utilize the point cloud data without additional bounding boxes to represent where an asset is located. This will change the industry as we know it. The opportunity to remove the requirement of expensive and time-consuming back modelling activities enables an expedited process to connect business process to a level of spatial fidelity that has not been achieved before. ?Providing simple spatial context, although not to an LOD500 geometric level of detail, however, will no doubt suffice for the asset owner and maintenance use cases that are currently being explored.

Asset Owner Vision Statements

1. Spatial Digital Thread – Enabling system integration through the use of space, location and time.
2. Spatial Data Management for Asset Owners with a focus of high fidelity change over time.

With all this said, the concept of unsupervised classification and segmentation AI / ML for point cloud data indexing is some way off. However, we need to start somewhere. Grouped point cloud data won’t answer all spatial business requirements, and a hybrid of multiple spatial nodes will still be required. This is where classification becomes important. Instead of defining standards based on metadata, lets base industry standards on spatial level of detail to achieve some form of business value. For example, the classification below will enable a more cost-effective way to manage various spatial nodes based on what the asset owner determines to be valuable to their operating model based on asset class. ?

Classification is important from a facility or area reporting perspective, however the key to sustainable success for spatial data for the asset owner is not BIM standards or extensive classification, but object instance segmentation of point cloud data and the ability to define a relationship between the segmented space and objects that are contained within. This will allow the concept of the spatial database to be created and a simple way to relate all spatial instances to each other via the spatial index they consume. This will need to be multiresolution in structure and dynamic in its model tree to ensure multithreaded complex relationships can be created based on the user requirements and use case.?