Courtesy of Acute3D
Photo Magic Image frames captured from multiple vantage points are automatically analyzed for common points to generate 3D models.

Reality capture has been a foundation of design and construction since the age of chains and transits. Now, the technology is rapidly improving in sophistication, automation and integration.

In one of the latest examples, Bentley Systems Inc., Exton, Pa., on Feb. 6 acquired a four-year-old French company, Acute3D, whose Smart3DCapture software already has an international clientèle for the automated transformation of batches of digital photos into 3D models up to city scale.

Bentley plans to integrate the capability in its infrastructure design, construction and support products to make reality capture and 3D model creation a native capability.

"Thanks to the integration within Bentley, we will be able to provide more vertical solutions, which will be useful for the entire infrastructure life cycle," says Jean-Phillipe Pons, CEO and co-founder and, with Renaud Keriven, of Acute3D prior to the acquisition. Both are now software development directors with Bentley.

"We have a grand vision called reality modeling," says Styli Camateros, a senior vice president for Bentley's infrastructure products and civil and geospatial lines. He notes that devices, such as cameras and drones, that are capable of capturing photos and accurate geometric data about physical objects are proliferating "in a big way," and Acute3D will let users "very quickly" convert their images into 3D models. "Not only is their product great, but we intend to put it in all of our products," Camateros says.

Smart3DCapture produces 3D models from photos by analyzing images of a static subject taken from different viewpoints and identifying pixels in the separate images that correspond to the same point in a scene. With enough point correlations, an accurate 3D shape can be inferred.

Joseph Seppi, national security program manager of Dayton, Ohio-based Woolpert, Inc., an architecture, engineering and geospatial firm, says his company has been using Acute3D since last summer and finds it works particularly well with extremely large image sets. “There are some other products out there, but none that work as well or on large projects involving tens of thousands, or even hundreds of thousands of input images,” Seppi says.

“We are using it with airborne data primarily. We are collecting pretty large areas, the size of a city for example, and taking all of that imagery and running it though Acute3D. It will actually create a very dense, colorized point cloud up to the average native image resolution, or you can do this textured mesh; that’s what they call their model. It does not contain any intelligence about the features. It’s not doing auto feature extraction, its not classifying anything, its just doing a very accurate 3D reconstruction of the objects in a scene.” Seppi says the output can be taken into a classification package for further feature extraction.

“What’s special about it is that the quality of the models is quite good. We have looked at some other products and open source tools. They either don’t do as good a job, or they do as good a job but can’t handle the amount of data,” Seppi says.

Bentley's Camateros says "We think it is a core functionality that will bring a lot of value."

In a statement about the acquisition, Keith Bentley, Bentley Systems founder and CTO, observed, "Acute3D has made it possible for anyone to sufficiently capture existing conditions with just a camera. … [But] rather than a voluminous cloud of discrete points, Acute3D produces a 3D 'reality mesh' intrinsically in the same geometric idiom as engineering models, readily aligning the real-world context."