Mobile 3D indoor mapping using the Continuous Normal Distributions Transform

Dylan Campbell, Mark Albert Whitty, Samsung Lim

Abstract: Existing approaches for indoor mapping are often either time-consuming or inaccurate. This paper presents the Continuous Normal Distributions Transform (C-NDT), an efficient approach to 3D indoor mapping that balances acquisition time, completeness and accuracy by registering scans acquired from a rotating LiDAR sensor mounted on a moving vehicle. C-NDT uses the robust Normal Distributions Transform (NDT) algorithm for scan registration, ensuring that the mapping is independent of the long-term quality of the odometry. We demonstrate that C-NDT produces more accurate maps than stand-alone dead-reckoning, achieves better map completeness than static scanning and is at least an order of magnitude faster than existing static scanning methods.