Breakthrough in Multi-Floor SLAM Algorithms: GreenValley Empowers Efficient Property Surveying
-- 03 Sep 2025 --
Background
In recent years, Simultaneous Localization and Mapping (SLAM) technology for 3D laser scanning has advanced significantly. It enables the generation of high-precision point cloud data in indoor environments without GNSS signals, accurately reconstructing building models. SLAM technology has unique advantages in property surveying, making it increasingly prominent in the field. Studies have shown that the precision of data from SLAM 3D laser scanners meets the requirements for property surveying, achieving an efficiency over five times that of traditional surveying methods. This technology and related equipment hold great potential in property surveying applications.
Industry Challenges
On the one hand, although various derivative SLAM technologies, such as visual SLAM, have been developed to improve point cloud accuracy, the principles of typical visual sensors limit their application in close-range scenes, while monocular SLAM often suffers from scale drift [1].
On the other hand, while the angular resolution of LiDAR sensors decreases with distance due to diffraction, the scanning effect largely depends on the scanning pattern and the line density defined by the sensor's spatial resolution. Errors accumulate as data collection time increases and rapid turns during data collection can lead to data drift and feature trajectory loss, causing point cloud misalignment and tilting.
As a result, SLAM devices often encounter tilting and misalignment issues when scanning multi-floor buildings, significantly limiting their application in such environments. Current solutions mainly involve re-scanning and stitching data for every two to three floors, which is time-consuming and prone to accuracy issues influenced by human factors during outdoor data collection and stitching processes.
Before Stitching (Left) and After Stitching (Right)
Solution
With GreenValley's LiDAR360MLS 8.0 multi-floor algorithm, there is no need for multiple project stitching. Multi-floor data can be scanned and processed within a single project, achieving both speed and precision.
The new SLAM algorithm incorporates parallelism and verticality considerations, significantly enhancing spatial feature recognition and optimization. By integrating supplementary observations such as IMU data, it improves the reliability and accuracy of SLAM data processing.
To test the effectiveness of the new algorithm, a client conducted data collection for a 12-floor building, completing the task in just 31 minutes using GreenValley's in-house-developed handheld SLAM 3D laser scanner, the LiGrip H300. Below is an example of the results obtained after scanning and data processing.
Below: Raw data processed with LiDAR360MLS 8.0 software. Profiles from both facade and side views show an average parallelism of approximately 0.031°.
Conclusion
The multi-floor SLAM algorithm proposed by GreenValley has been widely applied in outdoor surveying of complex multi-floor buildings. This algorithm demonstrates exceptional accuracy in parallelism for multi-floor structures, especially in challenging indoor scenarios like narrow corridors and staircases. It effectively addresses the time and labor-intensive challenges of multi-floor building surveying. As 3D laser scanning technology and post-processing software continue to advance, the application prospects of SLAM 3D laser scanners in property surveying will become even broader.
References
[1] Vidanapathirana, K., Ramezani, M., Moghadam, P., Sridharan, S., Fookes, C., 2022. Logg3d-net: Locally Guided Global Descriptor Learning for 3D Place Recognition. 2022 International Conference on Robotics and Automation (ICRA), IEEE, 2215–2221.
Definitions
Parallelism: The degree of alignment between building facades and floor surfaces.
Verticality: The degree of alignment between building cross-sections and wall surfaces.
In recent years, Simultaneous Localization and Mapping (SLAM) technology for 3D laser scanning has advanced significantly. It enables the generation of high-precision point cloud data in indoor environments without GNSS signals, accurately reconstructing building models. SLAM technology has unique advantages in property surveying, making it increasingly prominent in the field. Studies have shown that the precision of data from SLAM 3D laser scanners meets the requirements for property surveying, achieving an efficiency over five times that of traditional surveying methods. This technology and related equipment hold great potential in property surveying applications.
Industry Challenges
On the one hand, although various derivative SLAM technologies, such as visual SLAM, have been developed to improve point cloud accuracy, the principles of typical visual sensors limit their application in close-range scenes, while monocular SLAM often suffers from scale drift [1].
On the other hand, while the angular resolution of LiDAR sensors decreases with distance due to diffraction, the scanning effect largely depends on the scanning pattern and the line density defined by the sensor's spatial resolution. Errors accumulate as data collection time increases and rapid turns during data collection can lead to data drift and feature trajectory loss, causing point cloud misalignment and tilting.
As a result, SLAM devices often encounter tilting and misalignment issues when scanning multi-floor buildings, significantly limiting their application in such environments. Current solutions mainly involve re-scanning and stitching data for every two to three floors, which is time-consuming and prone to accuracy issues influenced by human factors during outdoor data collection and stitching processes.
Before Stitching (Left) and After Stitching (Right)
Solution
With GreenValley's LiDAR360MLS 8.0 multi-floor algorithm, there is no need for multiple project stitching. Multi-floor data can be scanned and processed within a single project, achieving both speed and precision.
The new SLAM algorithm incorporates parallelism and verticality considerations, significantly enhancing spatial feature recognition and optimization. By integrating supplementary observations such as IMU data, it improves the reliability and accuracy of SLAM data processing.
To test the effectiveness of the new algorithm, a client conducted data collection for a 12-floor building, completing the task in just 31 minutes using GreenValley's in-house-developed handheld SLAM 3D laser scanner, the LiGrip H300. Below is an example of the results obtained after scanning and data processing.
Below: Raw data processed with LiDAR360MLS 8.0 software. Profiles from both facade and side views show an average parallelism of approximately 0.031°.
Conclusion
The multi-floor SLAM algorithm proposed by GreenValley has been widely applied in outdoor surveying of complex multi-floor buildings. This algorithm demonstrates exceptional accuracy in parallelism for multi-floor structures, especially in challenging indoor scenarios like narrow corridors and staircases. It effectively addresses the time and labor-intensive challenges of multi-floor building surveying. As 3D laser scanning technology and post-processing software continue to advance, the application prospects of SLAM 3D laser scanners in property surveying will become even broader.
References
[1] Vidanapathirana, K., Ramezani, M., Moghadam, P., Sridharan, S., Fookes, C., 2022. Logg3d-net: Locally Guided Global Descriptor Learning for 3D Place Recognition. 2022 International Conference on Robotics and Automation (ICRA), IEEE, 2215–2221.
Definitions
Parallelism: The degree of alignment between building facades and floor surfaces.
Verticality: The degree of alignment between building cross-sections and wall surfaces.
