Experimental assessment of IMU data processing techniques for a backpack mobile laser scanning system

Abstract

Positioning techniques are fundamental in many automation tasks with several applications. In GNSS-denied environments like in dense forests, other alternatives are required, such as inertial and visual navigation. However, Inertial Measurement Units (IMUs) data, mainly those from microelectromechanical-system (MEMS), are noisy, which affects the orientation estimation. MEMS IMUs have been employed in mobile laser scanning systems due to their compact design and low-cost solutions for short-term navigation. In this paper, we have compared three IMU processing techniques freely available: MAH (Mahony et al., 2009), MAD (Madgwick et al., 2011) and DCM (Hyyti and Visala, 2015). These techniques implemented different approaches to estimate the attitude. They were experimentally assessed with data from a backpack mobile laser scanning system, which is composed of an OS0-128 Ouster LiDAR equipped with an internal IMU. We have used data from a 5-second trajectory segment aiming to evaluate the attitude and position estimation for a local path. The results showed that the DCM algorithm maintained a consistent velocity for 5 seconds, achieving a positional error of 1.4 m, 0.06 m, and 1.05 m along the X-, Y- and Z-axis, respectively. In contrast, MAD and MAH showed a position error over 20 m, 7 m and 3 m along the X-, Y- and Z-axis, respectively, which was affected by the velocity drift.