Svein Ivar Vestbø has been scanning and producing 3D point clouds of structures for years. A survey engineer with Norway’s Geomatikk Survey, his assignments usually include residential houses or low-rise commercial units for architecture firms and other organizations. That all changed about a year ago when the company took on an unusual restoration project that would take their scanning and surveying expertise to new heights.
The site was the 150-year-old former home of Bergens Tidende, Norway’s daily regional newspaper, located in Bergen’s city center. Measuring 2,560 square meters and spanning a city block, the building is now home to the University of Bergen's social science center, photo archives and other UiB offices. To meet UiB’s need for a more functional working environment, the university decided to reconstruct the interior of the structure. Capturing such a massive building would be a significant challenge in scanning and data processing.
“The scope of this work was intense,” says Vestbø. “The architect needed a 3D model of the building’s façade at a precision of 2 millimeters, which would require a huge number of scans. And the building has more than 500 windows and other glass features. That means the scanner would shoot through windows and collect millions of extraneous data points inside the building, which would increase complexity in data processing. Our measurements of the height of the multiple floors inside needed to be within 0.5 mm, which is incredibly tight.”
Such a tall order required precise, efficient scanning and surveying technology, robust data-processing software, and an innovative data-integration approach. Vestbø and his team leveraged these three skill sets and delivered the 3D data depth and breadth needed to help modernize and transform the historic building.
The architecture firm, Artec, awarded Geomatikk Survey the BT contract in mid-December 2018. It was a three-phased project, and company surveyors would need to scan the building’s facade, roof, courtyards and backyards before measuring and calculating varying floor heights so they could scan the existing walls and spaces after the interior was demolished. The BT building is a boxy structure that measures 26 meters high, 64 meters long and 40 meters wide and contains 10 floors. In overall footprint, that equates to about two Olympic-sized swimming pools.
Five days after winning the project, Vestbø was ready to begin phase one.
With the risk of inclement weather, Vestbø decided to capture the roof first. Viewed from above, the building resembles a thick W: the flat roof has two long rectangular ends running vertically, a wide horizontal strip across the length of the building and an elevated rectangular block in the middle. Dotted around the surfaces are various utility structures and a cluster of picnic tables and benches.
The SX10 atop the UiB building, with a clear view of Bergen. Fortunately, not a snowflake fell during the scanning of the building. Photo credit: Geomatikk Survey
Using a Trimble R10 GNSS receiver, Vestbø established site control by setting three temporary fixed points on the roof. He observed each point for about one minute, waited 15 minutes, and then repeated the process two more times. Tying into Norway’s CPOS GNSS measurement correction service, he obtained an initial overall accuracy between 8 to14 millimeters in plane view and 2 to 3 centimeters accuracy in vertical. They saved the R10 data using Trimble Access™ field software on a Trimble TSC3 data controller.
Vestbø occupied each temporary fixed point with the total station mode of the Trimble SX10 scanning total station and measured those points. He then moved the instrument to two other roof locations to establish a set of new benchmarks for a tighter control network. Integrating the R10 and SX10 measurements created an average accuracy of 3 millimeters in horizontal, vertical and elevation.
From high to low
Per project specifications, Geomatikk Survey needed to scan the BT building’s facade, roof, courtyards, and backyard features and produce a point cloud with an overall accuracy of 2 millimeters. In addition, the architect required a second point cloud that would include the building’s surroundings, including trees and buildings across the street.
To capture the entire building and meet that precision, Vestbø established seven georeferenced scan locations, including roof corners, street-level building corners and the back of the building, and set up an additional 45 scan stations; those non-registered scans would be tied into the georeferenced scans in post processing.
Setting the SX10 on the roof positions, Vestbø first scanned all the roof’s features including picnic tables and benches, utilities, raised structures, air vents and fencing along the external edges. From that vantage point, he also had clear sight to scan the courtyards and backyard features, including a glassed-roof atrium, 30 meters below. He then moved to the street, and methodically moved around the BT structure, setting up and leveling the SX10 and capturing a full dome scan of every detail of the facade, as well as the courtyards and backyards.
Per project specifications, Vestbø created a secondary point cloud showing trees and other buildings surrounding the BT building.
At scanning distances ranging from 2 to 30 meters, each scan took 12 minutes to complete. Vestbø then spent another three minutes using the scanning total station’s built-in camera to take a series of photos. After 13 hours of scanning, Vestbø had collected 260 million data points with a 40-60 percent overlap.
“The scanner is surprisingly quick and quite precise,” Vestbø said. “And with so many overlapping points, it makes it much easier to process the data and create an accurate point cloud. It also automatically registers the data so you don’t have to georeference the point cloud during post-processing, which is a huge time savings.”
Automation plays key role
Back in the office, Vestbø used the drag-and-drop capability of Trimble Business Center software to integrate the GNSS and scanning data and colorize the scan. He then used the seven georeferenced scans to spatially anchor the 45 unregistered scans. He manually delineated a window of referenced points and the software automatically georeferenced each scan one by one to create a spatially accurate 3D point cloud –– a process that took about two minutes per scan.
Automation also played a key role in cleaning up the 3D dataset. Scanning through 550 windows and two different glass roofs created millions of extraneous points inside the building, but using TBC’s limit box tool, Vestbø could remove exterior walls to expose the interior data points. After setting certain parameters and boundaries, the software automatically removed 80 million points of additional data captured inside the building.
Three days after completing the phase one scanning fieldwork, Vestbø delivered the precise 3D model of the BT building facade and the secondary point cloud to the architect.
Leveling the floors
A few months later, Vestbø and a colleague returned to the site for phase two of the project: measuring and calculating the height of each of the 10 floors in the building. Because of the age of the building, the floors were not constructed with uniform heights. Some floors had slopes, or dips or raised sections, which created varying heights between floors. The Geomatikk Survey team needed to accurately survey and calculate the elevation levels of each floor to within 0.5 millimeters.
The crew first established a set of benchmarks for survey control. Vestbø placed a prism on the inside of a window on each floor. Setting up on the street, he used the SX10’s total station functionality to shoot and record the position of each prism.
Once inside, they used a digital leveling instrument to collect the height data. Because they weren’t allowed to draw on the floor, they used tape to mark measuring points. Using the established benchmark point as the starting point, they methodically moved along each floor, collecting measurements from the existing walls, doors and openings and recording, on average, 20 height measurements for each floor.
To add value to the height data points, Vestbø took 2D DWG files of the building that he acquired from UiB and used TBC to integrate them into the finished 2-millimeter point cloud. Two days after completing the leveling fieldwork, he provided the architect with an updated point cloud layered with the DWG floor drawings and the calculated heights on each floor.
“Scanning not only captures all the unique details of the building quickly, it provides so much data depth and feature definition that it’s like you’re standing in front of the building or inside it from your own computer,” Vestbø said. “A 3D model will enable the architects to better evaluate and visualize the structure and incorporate that precise view into their own design model and workflow.” That view will help preserve the integrity of the building while increasing its functionality for many years to come.
Hero Image:The colorized 2-mm point cloud of the BT building. Photo credit: Geomatikk Survey