Moving Humans Removal for Dynamic Environment Reconstruction from Slow-Scanning LIDAR Data

Point clouds data registration is a crucial task for robotic environment reconstruction and understanding. The dynamic environment is a challenging problem for point clouds registration since moving objects occlude the static background and result in the feature corresponding failure. In this paper, we propose a moving object detection and removing method for slow-scanning LIDAR data. We define a Mean Axis Descriptor for moving objects description. Moreover, we recover the big motion distortions using this descriptor and sensor parameters. These recovered objects can be used for motion tracking and high frame-rate point clouds generation. A big scale dynamic environment reconstruction experiment result indicates that the proposed method is promising in moving object removal and motion distortion recovery.