3D Deep Learning Classification Method for Airborne LiDAR Point Clouds Fusing Spectral Information

Hongtao Wang; Xiangda Lei; Zongze Zhao

doi:10.3788/LOP57.122802

Laser & Optoelectronics Progress, Volume. 57, Issue 12, 122802(2020)

3D Deep Learning Classification Method for Airborne LiDAR Point Clouds Fusing Spectral Information

Hongtao Wang, Xiangda Lei^*, and Zongze Zhao

Author Affiliations

School of Surveying and Land Information Engineering, Henan Polytechnic University, Jiaozuo, Henan 454000, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(10)

Fig. 1. Classification flow of 3D point clouds based on PointNet

Download full size

Fig. 2. Flow chart of classification for airborne LiDAR point clouds fusing multi-spectral imagery

Download full size

Fig. 3. Multi-scale grid processing on point cloud data

Download full size

Fig. 4. Training set and the corresponding multi-spectral imagery. (a) Training set; (b) multi-spectral imagery

Download full size

Fig. 5. Test set and the corresponding multi-spectral imagery. (a) Test set; (b) multi-spectral imagery

Download full size

Fig. 6. Classification results of original LiDAR point clouds. (a) Classification results of different objects; (b) misclassification results

Download full size

Fig. 7. Classification results of multi-spectral point clouds. (a) Classification results of different objects; (b) misclassification results

Download full size

Table 1. Comparison of classification accuracy under different scales
View table
Table 1. Comparison of classification accuracy under different scales
Size /m OA /% Kappa
2 74.88 0.6756
5 81.54 0.7592
10 81.14 0.7541
15 78.36 0.7178
2,5,10 80.00 0.7396
5,10,15 82.02 0.7658

Table 2. Point cloud classification results of unfused and fused spectral information
View table
Table 2. Point cloud classification results of unfused and fused spectral information
Type of data F₁ /% OA /% Kappa
Low vegetation Impervious surface Car Roof Shrub Tree
Original data 73.60 83.28 35.88 69.06 37.01 51.86 68.63 0.5969
Fused data 75.92 85.90 54.03 94.08 42.69 79.47 82.02 0.7658

Table 3. Accuracy comparison of different classification methods
View table
Table 3. Accuracy comparison of different classification methods
Method F₁ /%
Low vegetation Impervious surface Car Roof Shrub Tree
Ours 75.9 85.9 54.0 94.1 42.7 79.5
IIS_7 65.2 85.0 57.9 90.9 39.5 75.6
UM 79.0 89.1 47.7 92.0 40.9 77.9
NANJ 77.7 90.9 51.7 93.6 - 77.1

Tools

Get Citation

Copy Citation Text

Hongtao Wang, Xiangda Lei, Zongze Zhao. 3D Deep Learning Classification Method for Airborne LiDAR Point Clouds Fusing Spectral Information[J]. Laser & Optoelectronics Progress, 2020, 57(12): 122802

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites