Target Classification of Hyperspectral Lidar Based on Optimization Selection of Spatial-Spectral Features

Bowen Chen; Shuo Shi; Wei Gong; Qian Xu; Xingtao Tang; Sifu Bi; Biwu Chen

doi:10.3788/AOS221717

Journals >Acta Optica Sinica >Volume 43 >Issue 12 >Page 1228008 > Article

Acta Optica Sinica
Vol. 43, Issue 12, 1228008 (2023)

Target Classification of Hyperspectral Lidar Based on Optimization Selection of Spatial-Spectral Features

Bowen Chen^1、2、3, Shuo Shi^{2、3、4、*}, Wei Gong^2、3、4, Qian Xu², Xingtao Tang², Sifu Bi², and Biwu Chen⁵

Author Affiliations

¹Chinese Antarctic Center of Surveying and Mapping, Wuhan University, Wuhan 430079, Hubei, China

²State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079, Hubei, China

³Electronic Information School, Wuhan University, Wuhan 430079, Hubei, China

⁴Collaborative Innovation Center of Geospatial Technology, Wuhan 430079, Hubei, China

⁵Shanghai Radio Equipment Research Institute, Shanghai 201109, China

show less

DOI: 10.3788/AOS221717 Cite this Article Set citation alerts

Bowen Chen, Shuo Shi, Wei Gong, Qian Xu, Xingtao Tang, Sifu Bi, Biwu Chen. Target Classification of Hyperspectral Lidar Based on Optimization Selection of Spatial-Spectral Features[J]. Acta Optica Sinica, 2023, 43(12): 1228008 Copy Citation Text

show less

Fig. 1. System prototype of hyperspectral lidar

Download full size

Fig. 2. Scanning scene of 14 different targets

Download full size

Fig. 3. Target classification processes based on hyperspectral lidar for spatial-spectral feature optimization selection

Download full size

Fig. 4. True color reconstruction result based on optimal band combination

Download full size

Fig. 5. Real categories of fourteen targets

Download full size

Fig. 6. Target classification results of the first four classification strategies. (a) 32-channel spectral information and elevation value; (b) spectral indices; (c) geometric features; (d) spatial-spectral feature combination

Download full size

Fig. 7. Target classification results of the 5th and 6th classification strategies. (a) Spatial-spectral features selected by marine predator algorithm; (b) optimal spatial-spectral feature combination

Download full size

Fig. 8. Correlation of spatial-spectral features

Download full size

Vegetation index	Formula	Reference
Differential vegetation index（DVI）	$R_{775} - R_{675}$	［25］
Ratio vegetation index（RVI）	$\frac{R_{775}}{R_{675}}$	［26］
Enhanced vegetation index（EVI）	$\frac{2.5 (R_{775} - R_{675})}{1 + R_{775} + 2.4 R_{675}}$	［27］
Soil adjusted vegetation index（SAVI）	$\frac{1.5 (R_{775} - R_{675})}{R_{775} + R_{675} + 0.5}$	［28］
Normalized differential vegetation index（NDVI）	$\frac{R_{775} - R_{675}}{R_{775} + R_{675}}$	［29］
Ratio normalized differential vegetation index（RNDVI）	$\frac{R_{775}^{2} - R_{675}}{R_{775} + R_{675}^{2}}$	［30］
Red-edge chlorophyll index（CIred-edge）	$\frac{R_{745}}{R_{705}} - 1$	［31］
Modified chlorophyll absorption ratio index（MCARI）	$\frac{R_{695}}{R_{665}} - 0.2 (R_{695} - R_{495}) \frac{R_{695}}{R_{665}}$	［32］
Plant senescence reflectance index（PSRI）	$\frac{R_{675} - R_{495}}{R_{745}}$	［33］
Triangular vegetation index（TVI）	$0.5 [120 (R_{745} - R_{495}) - 200 (R_{665} - R_{545})]$	［34］

Table 1. Vegetation indices used in this study

Color index	Formula	Reference
Excess green index（ExG）	$2 G - R - B$	［38］
Normalized green‑red difference index（NGRDI）	$\frac{G - R}{G + R}$	［39］
Normalized green‑blue difference index（NGBDI）	$\frac{G - B}{G + B}$	［40］
Excess red index（ExR）	$1.4 R - G$	［41］
Excess green minus excess red（ExGR）	$E_{x} G - E_{x} R$	［42］
Visible atmospherically resistant index（VARI）	$\frac{G - R}{G + R - B}$	［43］
Visible‑band diference vegetation index（VDVI）	$\frac{2 G - R - B}{2 G + R + B}$	［44］
Modified green red vegetation index（MGRVI）	$\frac{G^{2} - R^{2}}{G^{2} + R^{2}}$	［45］
Red green blue vegetation index（RGBVI）	$\frac{G^{2} - R B}{G^{2} + R B}$	［45］
Normalized redness intensity（NRI）	$\frac{R}{G + R + B}$	［46］
Green minus red difference index（GMR DI）	$G - R$	［47］

Table 2. Color indices used in this study

Target classification strategy	Classification feature
1	Original spectral information and elevation
2	Spectral index features
3	Spatial features
4	Original spectral information，elevation，spectral index features，and spatial features
5	Spatial‑spectral features selected by marine predator algorithm
6	Optimal spatial-spectral feature combination

Table 3. Six different target classification strategies

Strategy	1		2		3		4		5		6
OA /%	91.49		90.73		89.56		95.57		96.66		97.13
AA /%	77.74		78.27		68.26		84.37		87.44		89.05
Kappa	0.8934		0.8837		0.8693		0.9380		0.9577		0.9642
Time /s	4.72（±0.32）		4.18（±0.45）		3.53（±0.22）		5.16（±0.47）		4.06（±0.18）		3.61（±0.17）
Class	P	R	P	R	P	R	P	R	P	R	P	R
1	0.98	0.96	0.98	0.95	1.00	0.99	1.00	1.00	1.00	1.00	1.00	1.00
2	0.90	0.91	0.88	0.87	0.84	0.75	0.92	0.92	0.92	0.97	0.93	0.96
3	0.82	0.79	0.75	0.80	0.68	0.76	0.83	0.88	0.89	0.91	0.91	0.92
4	0.93	0.91	0.93	0.91	0.88	0.85	0.97	0.93	0.98	0.93	0.99	0.95
5	0.81	0.88	0.84	0.88	0.92	0.93	0.91	0.93	0.94	0.95	0.91	0.97
6	0.93	0.88	0.77	0.83	0.87	0.79	0.96	0.93	0.99	0.97	0.99	0.99
7	0.88	0.86	0.90	0.84	0.73	0.70	0.95	0.91	0.97	0.96	0.98	0.95
8	0.73	0.77	0.77	0.76	0.36	0.59	0.79	0.80	0.83	0.82	0.88	0.91
9	0.90	0.93	0.90	0.93	0.87	0.92	0.94	0.96	0.94	0.97	0.97	0.97
10	0.69	0.71	0.57	0.65	0.67	0.67	0.82	0.94	0.88	0.94	0.91	0.99
11	0.67	0.90	0.77	0.81	0.33	0.51	0.86	0.88	0.86	0.89	0.89	0.86
12	0.93	0.87	0.90	0.91	0.91	0.90	0.96	0.95	0.97	0.95	0.98	0.94
13	0.33	1.00	0.40	0.75	0.20	0.33	0.40	0.86	0.40	0.86	0.47	0.88
14	0.40	0.84	0.62	0.82	0.29	0.44	0.65	0.85	0.67	0.92	0.65	0.92

Table 4. Classification accuracy summary

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information