Point Clouds Classification Algorithm Based on Cloth Filtering Algorithm and Improved Random Forest

Doudou Xue; Yinglei Cheng; Xiaosong Shi; Xianxiang Qin; Pei Wen

doi:10.3788/LOP57.221017

Laser & Optoelectronics Progress, Volume. 57, Issue 22, 221017(2020)

Point Clouds Classification Algorithm Based on Cloth Filtering Algorithm and Improved Random Forest

Doudou Xue^1、*, Yinglei Cheng¹, Xiaosong Shi¹, Xianxiang Qin¹, and Pei Wen^1,2

Author Affiliations

¹Information and Navigation College, Air Force Engineering University, Xi'an, Shaanxi 710077, China

²The 93575 Unit, Chengde, Hebei 067000, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(8)

Fig. 1. Flow chart of ICSF filtering algorithm

Download full size

Fig. 2. Feature grid diagram

Download full size

Fig. 3. Weighted weakly correlated random forest model

Download full size

Fig. 4. Point clouds elevation distribution. (a) Unnormalized point cloud data; (b) normalized point cloud data

Download full size

Fig. 5. Classification results of ground points. (a) Traditional CSF algorithm; (b) improved CSF algorithm

Download full size

Fig. 6. Classification results of different algorithms. (a) Correct classification results; (b) classification results of traditional RF algorithm; (c) classification results of SVM algorithm; (d) classification results of proposed algorithm

Download full size

Table 1. Comparison of experimental indicators of the two models

View table

Table 1. Comparison of experimental indicators of the two models

Item	Improved RF
Item	Ground	Building	Vegetation	Artificial	OA	K
Ground	0.94	0.00	0.00	0.06
Building	0.00	0.89	0.11	0.00	0.88	0.85
Vegetation	0.00	0.11	0.86	0.03
Artificial	0.07	0.03	0.09	0.81
Item	ICSF+Improved RF
Item	Ground	Building	Vegetation	Artificial	OA	K
Ground	0.96	0.00	0.00	0.04
Building	0.00	0.91	0.09	0.00	0.91	0.90
Vegetation	0.00	0.08	0.9	0.02
Artificial	0.00	0.03	0.10	0.87

Table 2. Classification accuracy of different feature combinations
View table
Table 2. Classification accuracy of different feature combinations
Model CA OA K Time /s
Ground Building Vegetation Artificial
Traditional RF 0.91 0.86 0.83 0.81 0.86 0.82 27.75
SVM 0.94 0.89 0.88 0.82 0.89 0.87 28.20
ICSF + Improved RF 0.96 0.90 0.91 0.87 0.91 0.90 23.56

Tools

Get Citation

Copy Citation Text

Doudou Xue, Yinglei Cheng, Xiaosong Shi, Xianxiang Qin, Pei Wen. Point Clouds Classification Algorithm Based on Cloth Filtering Algorithm and Improved Random Forest[J]. Laser & Optoelectronics Progress, 2020, 57(22): 221017

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites