The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume XL-4/W5
https://doi.org/10.5194/isprsarchives-XL-4-W5-1-2015
https://doi.org/10.5194/isprsarchives-XL-4-W5-1-2015
11 May 2015
 | 11 May 2015

PRECISE INDOOR LOCALIZATION FOR MOBILE LASER SCANNER

R. Kaijaluoto and A. Hyyppä

Keywords: Mobile Laser Scanning, Indoor localization, Simultaneous Localization and Mapping, Point Cloud

Abstract. Accurate 3D data is of high importance for indoor modeling for various applications in construction, engineering and cultural heritage documentation. For the lack of GNSS signals hampers use of kinematic platforms indoors, TLS is currently the most accurate and precise method for collecting such a data. Due to its static single view point data collection, excessive time and data redundancy are needed for integrity and coverage of data. However, localization methods with affordable scanners are used for solving mobile platform pose problem. The aim of this study was to investigate what level of trajectory accuracies can be achieved with high quality sensors and freely available state of the art planar SLAM algorithms, and how well this trajectory translates to a point cloud collected with a secondary scanner.

In this study high precision laser scanners were used with a novel way to combine the strengths of two SLAM algorithms into functional method for precise localization. We collected five datasets using Slammer platform with two laser scanners, and processed them with altogether 20 different parameter sets. The results were validated against TLS reference. The results show increasing scan frequency improves the trajectory, reaching 20 mm RMSE levels for the best performing parameter sets. Further analysis of the 3D point cloud showed good agreement with TLS reference with 17 mm positional RMSE. With precision scanners the obtained point cloud allows for high level of detail data for indoor modeling with accuracies close to TLS at best with vastly improved data collection efficiency.