Robust LiDAR-inertial Calibration System and Refined Normal Estimation Module for 3D Mapping

The LiDAR-inertial simultaneous localization and mapping (SLAM) technology has been increasingly used to extract crop phenotypic traits in precision agriculture, such as crop height, planting density, and plant morphology. For the LiDAR odometry with an inertial measurement unit (IMU), the initial extrinsic parameters can directly affect the accuracy of odometry and mapping. However, the accurate initial extrinsic parameters need calibrating instruments in each experiment, which require complex processes and high costs. In this article, to address these issues, a point-based update strategy is adopted to update the LiDAR state and align them with inertial navigation data, enabling more accurate estimation of LiDAR-inertial extrinsic parameters. Additionally, this article present a refined method for point cloud normal vector estimation, which involves incorporating iterative weighted principal component analysis (PCA) and modifying reference point positions, enhancing mapping accuracy. The experimental results obtained with the various datasets demonstrate that our proposed methods can accurately estimate LiDAR-inertial extrinsic parameters and positively improve the accuracy of 3D modeling and mapping in precision agriculture.