Semantic Point Cloud Mapping of LiDAR Based on Probabilistic Uncertainty Modeling for Autonomous Driving

Abstract

LiDAR-based Simultaneous Localization And Mapping (SLAM), which provides environmental information for autonomous vehicles by map building, is a major challenge for autonomous driving. In addition, the semantic information has been used for the LiDAR-based SLAM with the advent of deep neural network-based semantic segmentation algorithms. The semantic segmented point clouds provide a much greater range of functionality for autonomous vehicles than geometry alone, which can play an important role in the mapping step. However, due to the uncertainty of the semantic segmentation algorithms, the semantic segmented point clouds have limitations in being directly used for SLAM. In order to solve the limitations, this paper proposes a semantic segmentation-based LiDAR SLAM system considering the uncertainty of the semantic segmentation algorithms. The uncertainty is explicitly modeled by proposed probability models which are come from the data-driven approaches. Based on the probability models, this paper proposes semantic registration which calculates the transformation relationship of consecutive point clouds using semantic information with proposed probability models. Furthermore, the proposed probability models are used to determine the semantic class of the points when the multiple scans indicate different classes due to the uncertainty. The proposed framework is verified and evaluated by the KITTI dataset and outdoor environments. The experiment results show that the proposed semantic mapping framework reduces the errors of the mapping poses and eliminates the ambiguity of the semantic information of the generated semantic map.

Towards High-Definition Maps: a Framework Leveraging Semantic Segmentation to Improve NDT Map Compression and Descriptivity

“…In this work, we have shown that semantic information enables separate processes for each semantic label which results into more optimal clustering of point cloud data. Furthermore, in previous works semantic information has been used to improve positioning [11], [12], [17]. Moreover, semantic information enables the removal of unwanted dynamic objects from the map.…”

“…For semantic segmentation, SE-NDT uses PointNet [16] that is a pioneering solution of a point cloud segmentation network that consumes raw point cloud data without voxelization or rendering. Cho et al proposed that the uncertainty of semantic information could also be used in the registration task [17].…”

Generalized LOAM: LiDAR Odometry Estimation with Trainable Local Geometric Features

“…To achieve a more accurate and robust estimation, several modifications for semantic registration have been developed, including taking into account the uncertainty of the labels [14], [15], using multi-class semantic segmentation [16], [17], [18], [19], and adaptively changing the matching costs for the planar and line points [20]. However, these approaches cannot take full advantage of the geometric features of a shape because they classify points into discrete labels and thus cannot seamlessly fuse the data into continuous geometric features.…”

An adaptive scheme for degradation suppression in Lidar based SLAM

“…The matching results are optimized iteratively to calculate the correspondences. Cho et al (2020) proposed a semantic registration-based SLAM system. He calculated the transformation relationship of consecutive point clouds using semantic information with a probability model and achieved comparable results.…”