Bentley Bent on Making the Complex Simple

For example, look at the practical problem. How much dirt do I need to move or add to the site? What is needed to build my building? What and when is the end of life of the bridge?

The company not only has a box full of tools to get those answers, but also presents them in the language that the people with those kinds of problems and responsibilities can understand. Bentley has dozens of products, and if you're new to the company, you may need a glossary. Two examples are Reality Modeling and ContextCapture.

At its annual “Year in Infrastructure” conference, CEO and founder Greg Bentley made the case in his keynote that since Bentley Systems' acquisition and assimilation of ContextCapture software in 2015, the company will lead the proliferation of reality modeling across users’ projects. That should make it a major contributor to infrastructure project delivery and asset performance, according to Bentley. Reality modeling enables continuous capture of infrastructure assets, producing engineering-ready reality meshes and “enlivening” them for immersive interaction.

To date, reality modeling has processed digital photos acquired from ground-level and/or aerial photography—particularly from the increasing industrial use of unmanned aerial vehicles. Bentley announced a breakthrough in ContextCapture that he said will be available next month: Point clouds from laser scanning can now be combined with photos, as “hybrid inputs,” for reconstruction into a reality mesh. One of the advantages of ContextCapture is that it performs the filtering and convolution necessary to produce an engineering-ready mesh that is immediately consumable within architectural and engineering workflow sand is much more visually understandable and more manageable than the often bulky and unintelligent point clouds themselves. This technology is compelling in every circumstance, with its range of scanner or camera devices that can capture asset characteristics in a range of capture conditions.

It sounds simple but it's not. Photos taken from an airplane can give you a view otherwise not possible. However, aircraft, no matter the size, whether 747s or 12-inch drones, don't fly straight or level, no matter how good the pilot is. Wind, air pockets and closed-loop latency are the culprits, so the path they take can be as twisted as a DNA strand.

To correct for the airplane's motion and create an accurate 3D stereo map or surface mesh, you have to do what's known in photogrammetry as rectification. It’s a bunch of trigonometry and geometry, but doable if you know the math and have the time or the computer resources.

In laser-generated point clouds, it was the laser that shot photons at unsuspecting things. Those photons bounce off and comeback and are collected with photon sensors, just like what you have in your smartphone, only better, faster and much more expensive. The net result is an incomprehensible bundle of points, the “point cloud,” or an imagery dataset. Now the task is to extract useful information from that mess, and that's called convolution. The cloud is actually 4D data; it incorporates x, y and z coordinates and time. And that gives you highly accurate measurement data about the objects you navigated with your laser. Convoluting a cloud is even more mind-bending than photogrammetric rectification—and it's only possible with the onset of computers.

Bentley recognized the potential in this technology and, in 2011, acquired UK-based Pointools, stating that the acquisition was similar to an acquisition 10 years prior that transformed its geospatial capabilities. Pointools was a developer of point cloud technology, and Bentley declared it would make point cloud data a fundamental data type for AEC and weave the technology throughout its foundational products, which it has now done.

After you've run the convolutions, you end up with what's known as a feature map, and I've seen pseudo-colored feature maps with such astonishing high resolution they almost look like a 3D rendering.

Bentley has made some precise and clever developments and acquisitions over the years, not the least of which was MicroStation, the foundation for the company. In March 2015, Bentley also acquired Acute3D, the developers of Smart3DCapture software. Smart3DCapture was the engine in Autodesk's 123D Catch (formerly known as Photofly), a cloud service with several million users and now marketed as Autodesk's ReCap photo solution.

So, Bentley now had all the pieces to measure any existing structure, or any place where construction might take place. But like most software developed independently, the programs couldn't communicate with each other.

That’s where Bentley's R&D came in. The company's geniuses scattered around the world in France, Canada, the United States, India and China. Led by Greg Bentley, Bhupinder Singh and Francois Valois, the team worked on the issues and managed to integrate the photogrammetric engines of Acute3D and the point cloud engines from Pointools into a congruent set of algorithms that would produce a common file structure.

They've been testing it for over a year in their own labs and with select clients and now have announced it's ready for general use and consumption. The primary file format is 3MX, a Bentley proprietary format, but it can also produce AGI (Cesium), ESRI and Collada files. And, if you want beautiful pictures of its results, you can ship the file to Bentley’s rendering and animation package, LumenRT, which we'll discuss later.

Bentley has brought state-of-the-art high tech to the architecture, engineering and construction industries without frightening them. A recent survey by McKinsey & Company found that the construction industry is among the least digitized, possibly because its applications are so challenging and because the industry itself is very large and moves slowly when adopting new trends. Bentley has introduced solutions that show real and immediate returns on investment, and that's a language the construction industry and its clients should eventually be able to understand.