Georeferencing kinematic modeling and error correction of terrestrial laser scanner for 3D scene reconstruction

Publication Name

Automation in Construction

Abstract

Georeferencing means to transform the point cloud data collected by the terrestrial laser scanner (TLS) into a global coordinate system for three-dimensional (3D) scene reconstruction. However, the current georeferencing methods are mainly applicable for the TLS operating in the static mode; when the TLS is moving (i.e., in the kinematic working mode) during the 3D scene reconstruction, severe georeferencing error will be generated due to the kinematic measurement noise and geometric falsification offsets. To bridge this research gap, a new method based on the kinemics state constraints converted measuring Kalman Filter (KSC-CMKF) and the Levenberg Marquardt Wavelet neural network (LM-WNN) is proposed to correct the TLS georeferencing error in the kinematic operating mode (i.e., TLS kinematic georeferencing error). This new method can automatically correct the georeferencing error and simultaneously reduce the TLS systematic bias to save time and cost for the kinematic georeferencing. In order to achieve the georeferencing error correction, two kinematic models are established to respectively address the straight-line movement and turn movement scenarios. Then, the LM-WNN algorithm is used to estimate the uncertain propagation of the TLS kinemics error model to optimize the kinematic parameters. Subsequently, the KSC-CMKF algorithm is applied to correcting the coordinate transformation error between the TLS spherical coordinate and the global Cartesian coordinate. Lastly, the pre-survey signalized spherical targets are adopted as the ground control points to evaluate the performance of the TLS kinematic georeferencing. Experimental results demonstrate that the overall accuracy of the proposed TLS kinematic georeferencing method for the 3D scene reconstruction is significantly improved and is superior to that of two existing popular methods.

Open Access Status

This publication is not available as open access

Volume

126

Article Number

103673

Funding Number

51974290

Funding Sponsor

National Natural Science Foundation of China

Share

COinS
 

Link to publisher version (DOI)

http://dx.doi.org/10.1016/j.autcon.2021.103673