A Closed-Form Dual Quaternion Model for Drift Correction in TLS Pose-Circuits

Area of Science:

• Geomatics Engineering
• Computer Vision
• Robotics

Background:

• 3D point cloud data acquisition via laser scanning is rapid but prone to registration errors.
• Accumulated errors lead to drift and global inconsistencies in 3D reconstructions.
• Existing drift correction models distribute errors along closed trajectories.

Purpose of the Study:

• To present a novel drift correction model for 3D point clouds.
• To address global inconsistencies caused by registration errors in laser scanning.
• To improve the accuracy and efficiency of 3D reconstruction in closed trajectories.

Main Methods:

• Development of a drift correction model based on the linear interpolation of dual quaternions.
• Simultaneous refinement of rotations and translations in closed trajectories.
• Avoidance of iterative computations and matrix decomposition for efficiency.

Main Results:

• Experimental evaluations on eight terrestrial laser scanning (TLS) datasets.
• Robust average error reduction of 26% in 3D reconstructions.
• Maximum error reduction of 41% observed in circuits with significant drift.

Conclusions:

• The proposed dual quaternion-based model offers an efficient and fast solution for pose-circuit correction.
• The method significantly improves pose accuracy in closed trajectories, reducing global inconsistencies.
• The sensor-agnostic nature allows application to various 3D mapping systems beyond TLS.