Cascaded Point Cloud Segmentation Algorithm for Unstructured Environments Based on LiDAR for Autonomous Vehicles

Xiujian Yang; Jialong Huang; Shengbin Zhang; Haicheng Xiao

doi:10.3788/AOS250431

Acta Optica Sinica, Volume. 45, Issue 8, 0815004(2025)

Cascaded Point Cloud Segmentation Algorithm for Unstructured Environments Based on LiDAR for Autonomous Vehicles

Xiujian Yang, Jialong Huang, Shengbin Zhang^*, and Haicheng Xiao

Author Affiliations

Faculty of Transportation Engineering, Kunming University of Science and Technology, Kunming 650500, Yunnan , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(16)

Fig. 1. Schematic diagram of the cascaded point cloud segmentation algorithm principle

Download full size

Fig. 2. Schematic diagram of point cloud distortion correction

Download full size

Fig. 3. Point cloud data distribution and grid model division. (a) Uniform polar coordinate grid; (b) point cloud quantity distribution; (c) DCZM model

Download full size

Fig. 4. Schematic diagram of local ground estimation iteration. (a) Initial ground; (b) the first iteration; (c) the second iteration; (d) the third iteration

Download full size

Fig. 5. Top-down schematic diagram of 3D voxel clustering based on spherical coordinates. (a) Original point cloud; (b) Hash table construction; (c) neighboring point search of adjacent voxels; (d) point cloud clustering

Download full size

Fig. 6. 3D point cloud clustering process based on eigenvalue and normal vector angle constraints

Download full size

Fig. 7. All-terrain autonomous vehicle

Download full size

Fig. 8. Experimental environment for the all-terrain autonomous vehicle

Download full size

Fig. 9. Comparison test of ground point cloud segmentation on the Semantic KITTI dataset. (a) LineFit algorithm; (b) GPF algorithm; (c) Patchwork algorithm; (d) Patchwork++ algorithm; (e) proposed algorithm

Download full size

Fig. 10. Comparison test of non-ground point cloud segmentation on the Semantic KITTI dataset. (a) Adaptive_Clustering algorithm; (b) CVC algorithm; (c) Travel algorithm; (d) proposed algorithm

Download full size

Fig. 11. Ground point cloud segmentation algorithm test in real-world environment. (a) Front camera image of the all-terrain autonomous vehicle; (b) LineFit algorithm; (c) GPF algorithm; (d) Patchwork algorithm; (e) Patchwork++ algorithm; (f) proposed algorithm

Download full size

Fig. 12. Above-ground object segmentation algorithm test in real-world environment. (a) Front camera image of the all-terrain autonomous vehicle; (b) Adaptive_Clustering algorithm; (c) CVC algorithm; (d) Travel algorithm; (e) proposed algorithm

Download full size

Table 1. Ground point cloud segmentation algorithm

View table

Table 1. Ground point cloud segmentation algorithm

Input： cloud_points
Output： ground_points， nonground_points
1 DCZM （cloud_points）
2 for i=0 to num_patches do in parallel
3 for i=r.begin（） to r.end（） do
4 extract_initial_seeds_（zone_idx， src， ground_tmp）
5 for i=0 to num_iter_-1 do
6 estimate_plane_（ground_tmp， feat）
7 ground_tmp. clear（）；
8 initialize a matrix points（N， 3）
9 for j=0 to N-1 do
10 set points（j，：）←src［j］.getV ector3fMap（）
11 result ← points ∗ feat.normal_
12 for r=0 to N-1 do
13 if i< num_iter_-1 then
14 if result［r］ < feat.th_dist_d_ then
15 add src［r］ to ground_tmp
16 else
17 if result［r］ < feat.th_dist_d_ then
18 add src［r］ to ground_points
19 else
20 add src［r］ to nonground_points

Table 2. Key parameters of the proposed method
View table
Table 2. Key parameters of the proposed method
Notation $N_{p, m}$
$τ_{d}$ /
m
$Δ ρ$ /
m
$Δ θ$ /
rad
$Δ ϕ$ /
rad
$θ_{t h}$ /
rad
$w$
Value ${600,100,50,30}$ 0.125 0.7 0.35 1.4 0.262 0.65

Table 3. Evaluation of ground point cloud segmentation

View table

Table 3. Evaluation of ground point cloud segmentation

Sequence	Method	Precision /%	Recall /%	Accuracy /%	F1 score	Time /ms
01	LineFit	98.51±0.1222	75.22±0.5787	84.95±0.4124	0.85±0.0039	198.51±2.3533
	GPF	90.63±0.3971	67.98±1.4908	76.40±1.1810	0.74±0.01251	27.25±0.5959
	Patchwork	92.95±0.4780	91.50±0.2160	89.44±0.2097	0.92±0.0016	10.31±0.2263
	Patchwork++	96.88±0.1709	87.36±0.3129	91.64±0.2467	0.91±0.0021	22.68±0.56
	Our	95.71±0.1873	89.43±0.2623	90.01±0.2102	0.93±0.0018	9.57±0.2145
09	LineFit	97.11±0.1430	81.94±0.7478	90.41±0.5097	0.89±0.0046	195.94±3.3923
	GPF	94.23±1.354	88.71±0.4316	92.27±0.3033	0.91±0.030	24.11±0.2977
	Patchwork	89.44±4.1241	94.16±0.1676	91.70±0.1122	0.92±0.0011	8.42±0.0397
	Patchwork++	90.74±0.1870	93.95±0.1391	93.05±0.9303	0.92±0.0010	17.67±0.3215
	Our	92.62±0.1260	92.63±0.1845	92.71±0.1046	0.93±0.0090	5.30±0.0614
10	LineFit	96.26±0.7859	79.28±0.5136	92.44±0.3544	0.87±0.0033	160.61±2.7992
	GPF	87.24±0.3049	87.36±0.4000	92.64±0.2578	0.87±0.0027	27.41±0.5875
	Patchwork	84.65±0.2653	90.34±0.3720	87.17±0.3237	0.88±0.0026	8.62±0.3977
	Patchwork++	83.82±0.4698	93.22±0.2057	92.98±0.2324	0.88±0.0031	15.05±0.1612
	Our	91.34±0.1523	89.08±0.3571	89.54±0.3213	0.90±0.0022	5.83±0.0735

Table 4. Evaluation of above-ground object segmentation

View table

Table 4. Evaluation of above-ground object segmentation

Sequence	Method	Segmentation entropy	Scene accuracy	Time /ms
01	Adaptive_Clustering	0.0152±0.0013	0.3410±0.0092	11.97±0.5236
	CVC	0.0667±0.0010	0.8771±0.0060	49.77±0.8310
	Travel	0.1542±0.0073	0.6212±0.0103	23.98±0.6069
	Our	0.1104±0.0017	0.9037±0.0036	19.48±0.2563
09	Adaptive_Clustering	0.0087±0.0005	0.6365±0.0130	108.80±1.6532
	CVC	0.2088±0.0149	0.8457±0.0067	31.97±0.7346
	Travel	0.1276±0.0025	0.6593±0.0066	32.68±0.5571
	Our	0.1521±0.0036	0.8900±0.0046	20.31±0.4174
10	Adaptive_Clustering	0.0160±0.0002	0.6304±0.0054	13.93±0.4072
	CVC	0.1235±0.0034	0.7873±0.0048	102.04±0.8200
	Travel	0.2092±0.0052	0.6650±0.0067	97.85±0.8225
	Our	0.1177±0.0011	0.8421±0.0049	72.46±0.7737

Tools

Get Citation

Copy Citation Text

Xiujian Yang, Jialong Huang, Shengbin Zhang, Haicheng Xiao. Cascaded Point Cloud Segmentation Algorithm for Unstructured Environments Based on LiDAR for Autonomous Vehicles[J]. Acta Optica Sinica, 2025, 45(8): 0815004

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites