Industries / Construction and property / Engineering and architecture / Surveying
Laser scanning
Laser scanning measures buildings, structures and sites from the ground by sweeping a laser across them and recording millions of points, which together form a point cloud, a 3D copy of every surface the beam reached.
6
Milestones achieved, from 1998 to 2019.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1998
A slow start for scanning
Cyra Technologies, an American start-up led by Ben Kacyra with Jerry Dimsdale as chief technologist, was selling its first-generation Cyrax 2400 scanner and CGP software by 1998. It sold only 50 systems in 1998 and 1999 together, by the account of Geoff Jacobs, who joined Cyra in 1998 in a senior marketing post. Leica Geosystems, the Swiss survey-instrument maker, took a minority stake in March 2000, and its own history lists the purchase of Cyra under 2000, saying "the age of high-definition 3D scanning starts."
Lessonthe instrument worked before the market was ready, and a survey-instrument maker's backing carried it until it was.
Credit
Ben Kacyra, Jerry Dimsdale, Cyra Technologies; Leica Geosystems
2003
Heritage recorded before loss
Ben Kacyra, founder of Cyra Technologies, founded CyArk in 2003 to record, archive and share heritage sites in 3D, as accurate digital records made before the sites are damaged or lost to climate change, conflict or urban development.
Credit
Ben Kacyra; CyArk
2011
One file for every scanner
ASTM E2807-11 set out the E57 file format, which stores 3D points, their color and intensity, and the photographs a scanner takes, in a documented mix of binary data and XML. The format belongs to no scanner maker, and libE57, a free open-source library, reads and writes it, so a scan can move between programs from different companies. The libE57 project now asks ASTM to begin work on a version 2.0.
Credit
ASTM Committee E57 on 3D Imaging Systems
2013
A scanner on a spring
SLAM (simultaneous localization and mapping) lets a moving scanner work out where it is from what it scans, so a person can scan while walking through a building or a tunnel. M. Bosse, R. Zlot and P. Flick of CSIRO, Australia's national science agency, described Zebedee, a 3D range sensor mounted on a spring, in IEEE Transactions on Robotics in 2012. GeoSLAM, a company co-founded by CSIRO to sell it, launched the ZEB1 in 2013, which CSIRO calls the first handheld SLAM scanner.
Credit
CSIRO (Bosse, Zlot, Flick); GeoSLAM
2017
Scanners get a common test
ASTM E3125-17 is a test for tripod scanners that turn about two axes (spherical-coordinate scanners) and work at ranges of 2 to 150 m. The scanner measures targets such as spheres, the distances between pairs of target centers are compared with reference values, and the errors are judged against the maker's stated maximum permissible errors, so different instruments can be compared. NIST, the US measurement laboratory, led the work through an ASTM working group set up in 2013. The standard's scope is limited to this kind of scanner.
Credit
NIST Dimensional Metrology Group (Prem Rachakonda, Balasubramanian Muralikrishnan, Katharine Shilling, Geraldine Cheok, Daniel Sawyer); ASTM Committee E57
2019
A scan guides the rebuilding
In 2010 Andrew Tallon, an art historian at Vassar College, laser-scanned Notre-Dame cathedral in Paris to study its structure, especially its flying buttresses, and he gave a copy of the point cloud to the cathedral's chief architect, Benjamin Mouton, before Mouton retired in 2013. After the fire of 15 April 2019, Tallon's widow Marie Tallon made sure the restoration architects had his data. The chief architect, Philippe Villeneuve, said the scan let them rebuild the vaults "without any hesitation from a dimensional or formal standpoint."
Lessona scan made for research became the record for rebuilding because a copy was already in the owner's hands.
Credit
Andrew Tallon, Vassar College
Watch for
Survey grade from a backpack
Emesent says its GX1, which joins a SLAM laser scanner with RTK satellite positioning (positions corrected in real time from a nearby base) and cameras, and can be carried as a backpack, on a pole, on a vehicle or by hand, reaches 5 to 10 mm global accuracy, "independently validated", and can cut the time to survey a site by up to 95%. The launch announcement does not say who did the validation.
Achieved whena named outside party publishes a test of the GX1 against surveyed control points, with the error at each point.
Claimed by
Emesent, February 2026
Five millimeters, maker-tested
XGRIDS says its Lixel L3, which combines lidar, a 4K camera and RTK positioning and also produces Gaussian splats (3D scenes built from millions of small colored blobs that look like photographs), has 5 mm relative accuracy, "demonstrated against measured ground truth in a documented test", and about 2.5 cm absolute accuracy in real time. In a 2026 comparison written by staff of Pix4D, whose own phone-based system was among those tested, the earlier XGRIDS Lixel K1 showed maximum errors of about 10 to 15 cm at RTK-surveyed checkpoints.
Achieved whenan outside test of the L3 publishes its errors at surveyed checkpoints.
Claimed by
XGRIDS, September 2026
Open
01
One course for walking scanners
A permanent, published test course for handheld and backpack scanners, with a corridor, a stairwell and an outdoor loop marked by surveyed targets, and a method for scoring the scanners as they are sold, hardware and maker's software together. Several models are walked around the same course and scored the same way.
The briefCredit · unclaimed02
Bought scan-to-model tools, scored
A test of the commercial programs that turn a point cloud into a building model (scan-to-BIM), run on the public test buildings researchers already use and scored with the researchers' own code, plus the hours a person needs to finish each model.
The briefCredit · unclaimed
Open 01 · the brief
One course for walking scanners
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A permanent, published test course for handheld and backpack scanners, with a corridor, a stairwell and an outdoor loop marked by surveyed targets, and a method for scoring the scanners as they are sold, hardware and maker's software together. Several models are walked around the same course and scored the same way.
Why it is still open
ASTM E3125 covers only tripod scanners. The Hilti SLAM Challenges, held over several years up to 2023, score SLAM programs on data recorded with Hilti's own sensor rigs, not the commercial scanners surveyors buy. Accuracy claims for new walking scanners are their makers' own (see Watch for), and the comparisons that exist are one-off tests, often tied to a maker. A 2025 study of first- and second-generation handheld scanners by Teppati Losè and colleagues at the Politecnico di Torino used a test track at FARO's offices in Stuttgart, with two co-authors from FARO, which sells such scanners. A 2026 ISPRS comparison was written by Pix4D staff and found Pix4D's system best: under 3 cm RMSE with steady RTK, against maximum errors of about 10 to 15 cm for the XGRIDS and SHARE scanners. 3Dsurvey, a photogrammetry software company, reviewed seven handheld scanners against their makers' specifications and wrote that "independent real-world comparisons remain rare." No permanent course on which each new model is scored the same way was found.
astm.orgarxiv.orgpmc.ncbi.nlm.nih.govisprs-archives.copernicus.org3dsurvey.si
Built with what exists
A university building and grounds, used with permission; a total station and printed checkerboard targets; borrowed or rented scanners such as the FARO (GeoSLAM) ZEB, NavVis VLX, Emesent and XGRIDS units; export to E57; CloudCompare to measure each target in each scan.
First step
Lay out and survey 40 targets in one building, then walk the course with one scanner and measure the targets in its point cloud.
Done looks like
At least 40 surveyed targets on a course of about 500 m; at least four scanner models, each walked three times by two people; the error at every target and the error against distance walked published for each run, and the method sent to ASTM Committee E57.
Open 02 · the brief
Bought scan-to-model tools, scored
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A test of the commercial programs that turn a point cloud into a building model (scan-to-BIM), run on the public test buildings researchers already use and scored with the researchers' own code, plus the hours a person needs to finish each model.
Why it is still open
The Scan-to-BIM challenge at the CVPR computer vision conference, held since 2021 (its third round in 2023), scores research entries on whether wall, column and door endpoints land within 5, 10 and 20 cm, and the German Hospital dataset released in April 2026 adds point clouds with ground-truth building models. The tools firms buy are described only by their makers and resellers: Matterport's 2026 guide gives EdgeWise "hours to days for automatic element extraction, followed by manual validation and refinement," with no scores on a shared test.
Built with what exists
The CVPR Scan-to-BIM challenge's 3D data and evaluation code; the German Hospital (DeKH) dataset; EdgeWise and other scan-to-model programs on trial licenses, after checking that their terms allow published results; Autodesk Revit or the free Bonsai program to inspect models; a timer.
First step
Get a trial license for one tool and run it on one floor of the challenge's 3D training set.
Done looks like
For at least three commercial tools on at least two buildings, the challenge's precision and recall at 5, 10 and 20 cm for walls, columns and doors, and the person-hours of fixing needed to reach the reference model, shared with each maker before publication.