LiDAR Data Classification Method Based on High Recognition Compound Derivative Feature

Hui Bai; Fengbao Yang

doi:10.3788/LOP202158.0528001

Laser & Optoelectronics Progress, Volume. 58, Issue 5, 0528001(2021)

LiDAR Data Classification Method Based on High Recognition Compound Derivative Feature

Hui Bai and Fengbao Yang^*

Author Affiliations

College of Information and Communication Engineering, North University of China, Taiyuan , Shanxi 030051, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(13)

Fig. 1. Source feature images of airborne LiDAR data. (a) First echo height; (b) last echo height; (c) echo intensity

Download full size

Fig. 2. Derivative features of airborne LiDAR data. (a) Elevation difference; (b) NDVI; (c) GNDVI

Download full size

Fig. 3. Relationship between GNDVI and total chlorophyll concentration^[10]

Download full size

Fig. 4. Ridge trust allocation function

Download full size

Fig. 5. GRNDVI value and classification accuracy change curve. (a) Curves of classification accuracy changes of buildings and roads; (b) curves of classification accuracy changes of trees and grasslands; (c) average classification accuracy

Download full size

Fig. 6. Classification results of different vegetation indexes based on fuzzy DS. (a) Visible light images; (b) artificial data; (c) classification results based on NDVI and fuzzy DS; (d) classification results based on GNDVI and fuzzy DS; (e) classification results based on GRNDVI and basic DS; (f) classification results based on GRNDVI and fuzzy DS

Download full size

Table 1. Complementary sets of distinguishing features
View table
Table 1. Complementary sets of distinguishing features
Feature Class
A B
HD T
IN T
GRNDVI

Table 2. Comparison groups of different vegetation index characteristics experiments and different DS methods

View table

Table 2. Comparison groups of different vegetation index characteristics experiments and different DS methods

Method	Vegetation index characteristics	Other features	Classification method
Method 1	NDVI	DSM、IN、HD	Fuzzy DS
Method 2	GNDVI	DSM、IN、HD	Fuzzy DS
Method 3	GRNDVI	DSM、IN、HD	Basic DS
Proposed method	GRNDVI	DSM、IN、HD	Fuzzy DS

Table 3. Classification accuracy of data set 1
View table
Table 3. Classification accuracy of data set 1
Method Building Tree Grass Road Average value
Method 1 0.8261 0.8923 0.8866 0.8360 0.8578
Method 2 0.8506 0.8771 0.8522 0.8998 0.8688
Method 3 0.8723 0.6443 0.7990 0.9198 0.8141
Proposed method 0.8620 0.8910 0.8864 0.9087 0.8920

Table 4. Classification accuracy of data set 2
View table
Table 4. Classification accuracy of data set 2
Method Building Tree Grass Road Average value
Method 1 0.8932 0.7383 0.8679 0.8109 0.8413
Method 2 0.8921 0.7603 0.8804 0.8272 0.8361
Method 3 0.9070 0.7141 0.8771 0.8365 0.8442
Proposed method 0.9021 0.7537 0.8804 0.8238 0.8451

Table 5. Classification accuracy of data set 3
View table
Table 5. Classification accuracy of data set 3
Method Building Tree Grass Road Average value
Method 1 0.8779 0.8240 0.8592 0.8566 0.8625
Method 2 0.8757 0.8415 0.8711 0.8711 0.8598
Method 3 0.8776 0.7978 0.8632 0.8898 0.8613
Proposed method 0.8874 0.8667 0.8344 0.8691 0.8737

Table 6. Classification accuracy of data set 4
View table
Table 6. Classification accuracy of data set 4
Method Building Tree Grass Road Average value
Method 1 0.8787 0.7851 0.8535 0.8656 0.8570
Method 2 0.8805 0.8140 0.8626 0.8824 0.8556
Method 3 0.8811 0.7699 0.8602 0.8968 0.8560
Proposed method 0.8883 0.7998 0.8604 0.8791 0.8639

Table 7. Classification accuracy of data set 5
View table
Table 7. Classification accuracy of data set 5
Method Building Tree Grass Road Average value
Method 1 0.8912 0.8025 0.8755 0.8302 0.8570
Method 2 0.8934 0.8122 0.8849 0.8444 0.8577
Method 3 0.9037 0.7714 0.8768 0.8615 0.8544
Proposed method 0.9016 0.8083 0.8866 0.8412 0.8631

Tools

Get Citation

Copy Citation Text

Hui Bai, Fengbao Yang. LiDAR Data Classification Method Based on High Recognition Compound Derivative Feature[J]. Laser & Optoelectronics Progress, 2021, 58(5): 0528001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites