Ground Segmentation Algorithm for Sloped Terrain and Sparse LiDAR Point Cloud

Abstract

Distinguishing obstacles from ground is an essential step for common perception tasks such as object detection-and-tracking or occupancy grid maps. Typical approaches rely on plane fitting or local geometric features, but their performance is reduced in situations with sloped terrain or sparse data. Some works address these issues using Markov Random Fields and Belief Propagation, but these rely on local geometric features uniquely. This article presents a strategy for ground segmentation in LiDAR point clouds composed by two main steps: (i) First, an initial classification is performed dividing the points in small groups and analyzing geometric features between them. (ii) Then, this initial classification is used to model the surrounding ground height as a Markov Random Field, which is solved using the Loopy Belief Propagation algorithm. Points are finally classified comparing their height with the estimated ground height map. On one hand, using an initial estimation to model the Markov Random Field provides a better description of the scene than local geometric features commonly used alone. On the other hand, using a graph-based approach with message passing achieves better results than simpler filtering or enhancement techniques, since data propagation compensates sparse distributions of LiDAR point clouds. Experiments are conducted with two different sources of information: nuScenes's public dataset and an autonomous vehicle prototype. The estimation results are analyzed with respect to other methods, showing a good performance in a variety of situations.

Patchwork++: Fast and Robust Ground Segmentation Solving Partial Under-Segmentation Using 3D Point Cloud

“…Narksri et al [17] proposed a RANSAC-based ground segmentation with multiple regions and addressed the problem posed by sloped terrain. Jiménez et al [18] addressed the issue of local geometric variation of the ground by adopting the channel-based Markov random field.…”

“…As mentioned in Section I, various object clustering methods [18]- [21] include ground segmentation as a prior step for object recognition. Because ground points are not the region of interest in object clustering methods, the elimination of ground points enhances both computational efficiency and accuracy.…”

Road classification using 3D LiDAR sensor on vehicle

“…The captured scan data for some 3D-LiDAR points are organised at the same azimuth angle. This type of approach is commonly used in channel-based algorithms [14][15][16][17]. In [18], a channel-based algorithm was used after clustering the raw sensor data within a 2.5D grid.…”

“…Rieken et al proposed a channel-based ground classification combined with an area-based consistency constraint that can adapt to slope changes up to a predefined extent [15]. In [16], a channel-based approach combined with a multi-label Markov random field technique was proposed. This approach considers the contextual information of the points and can classify scenarios with uneven terrain.…”

A Robust Gaussian Process-Based LiDAR Ground Segmentation Algorithm for Autonomous Driving

“…The line-based methods mainly consider the scanning characteristics of LiDAR. In this method, the ground is divided into different segments according to the preset angle, and then each segment is divided into different small bins according to the distance [19][20][21][22][23][24][25][26][27][28][29][30]. By judging the spatial features or other features of the points in each bin, the reference ground height of each bin is obtained to establish the distinction between ground points and nonground points.…”