Automatic Multiple LiDAR Calibration Based on the Plane Features of Structured Environments

Abstract

The multiple light detection and ranging (LiDAR) system has been attached to numerous autonomous vehicles to reduce blind spot risks and increase measurement resolution. It is also important to accurately aligning multi-LiDAR data because misaligned sensor data can affect the localization and perception algorithms of LiDAR sensors. However, precise extrinsic parameter estimation is a challenging research area because the feature points in LiDAR data cannot be precisely aligned owing to the LiDAR resolution. In this paper, we proposed a targetless multi-LiDAR calibration method without any initial information. In the proposed approach, a random sample consensus (RANSAC)-based plane fitting technique was utilized for plane extraction, and cross-feature points were extracted from the polar view transform of segmented planes. To resolve the discontinuity problem of optimization, a nonlinear optimization method was utilized, and the experiments were performed within a structured environment. Overall, the experimental results showed that the proposed method yields comparable results with state-of-theart methods.

Fast Positioning Model and Systematic Error Calibration of Chang’E-3 Obstacle Avoidance Lidar for Soft Landing

“…Physical models model the errors of LiDAR based on physical parameters, such as ranging, angle measurement, and quadrature errors, while empirical models are based on empirical parameters [ 14 , 15 , 16 , 17 ]. To reduce the correlation among model parameters, these parameters can be treated as observations with prior information [ 18 , 19 , 20 ], or the strongly correlated parameters are grouped and calibrated step by step [ 21 , 22 ].…”