面向林地建图的激光惯导融合SLAM算法

The SLAM algorithm of lidar inertial navigation fusion can build a high-precision three-dimensional point cloud map and provide reliable data to acquire critical forest tree parameter information.The existing LIO-SAM is challenging to extract trunk features and filter ground points.We proposed F-LIO-SAM based on LIO-SAM to solve the problem.The algorithm proposed in this paper includes two innovations:a trunk feature extraction method using arc radius and a ground point filtering method using inertial navigation attitude information.The effects of the feature extraction method and the ground point removal method of the two algorithms are compared,and the effects of different tree densities on the mapping and positioning accuracy are analyzed.The experimental results show that the number of trunk feature points extracted by F-LIO-SAM is 50%higher than LIO-SAM.The average absolute error of the final single tree positioning accuracy is 25%lower than that of LIO-SAM,and the root means the square error is 24%.The research shows that the F-LIO-SAM proposed in this paper is a SLAM algorithm suitable for single tree location in an understory environment,which provides technical support for the acquisition of forest parameters in forest investigation.