Ground segmentation from 3D point cloud using features of scanning line segments
Cheng Ziyang, Ren Guoquan, Zhang Yin     
Department of Vehicle and Electrical Engineering, Army Engineering University, Shijiazhuang, Hebei 050003, China

Overview: The development of unmanned vehicles is very rapid, but most of the studies are based on the urban environment, while the ground segmentation in the complex environment still faces many challenges. The problems include: 1) in the bumpy terrain, the platform will have changes in pitch, roll and suspension; 2) the LiDAR points are unevenly distributed, such as the measurement points in the area close to the LiDAR are densely distributed relatively, while the distribution of measurement points in the area away from the LiDAR is sparse, which results in a large range of gaps between different scanning lines; 3) in the case of processing a few millions of points, the accuracy and real-time of the segmentation are difficult to balance. This article conducts research aiming at the problem of accurately segmenting the ground in real-time from 3D point cloud in various environments. Considering that the existing methods are complex, long time consuming, or selected features are not universal, a ground segmentation algorithm based on the features of scanning line segments is proposed. The algorithm first performs de-noising and pose correction on the 3D point cloud, then divides the scanning line according to the Euclidean distance and absolute height difference between adjacent points. Next, the characteristics of the adjacent line segments such as spacing, slope, and absolute height difference are analyzed. The maximum likelihood estimation is used to solve the feature threshold function, which improves the adaptability of threshold. Finally, comprehensively considering the undulating and inclined complex terrain, combining the distribution characteristics of the features of scanning line segments, the scanning line segments are marked as segments of flat ground, segments of slope and segments of obstacle by formulating the new horizontal and vertical classification strategies: firstly select the line segment with the smallest height from the scanning line closest to the radar origin and mark it as the initial ground scanning line segment. Then determine the line segments type in the scanning line closest to the radar origin horizontally and determine the segments type in other scanning lines vertically. This algorithm has been successfully applied to the unmanned ground platform. The effect of the actual engineering application indicates that, the features selected in this paper have high sensitivity and easy extraction, which are less affected by noise than single point features. The segmentation algorithm is highly efficient and robust, which can detect the ground stably and efficiently in structured road scene, wild undulating road scene and complex undulating scene. And the comparative test results of the algorithm in this paper with the local elevation estimation algorithm in Ref. [13] and the feature fusion algorithm in Ref. [14] show that the segmentation effect of this algorithm is superior to the other two algorithms in accuracy and time-consuming.

Supported by the National Defense Pre-Research Foundation of China (9140A09031715JB34001)