Hyperspectral Wave Band Selection Based on Golden Sine and Chaotic Spotted Hyena Optimizer Algorithm

Tianhong Dai; Chunxue Sun; Jianping Huang; Qiancheng Xie; Shijie Cong; Xinwang Huang; Kexin Li

doi:10.3788/LOP202259.1030003

Journals >Laser & Optoelectronics Progress >Volume 59 >Issue 10 >Page 1030003 > Article

Laser & Optoelectronics Progress
Vol. 59, Issue 10, 1030003 (2022)

Hyperspectral Wave Band Selection Based on Golden Sine and Chaotic Spotted Hyena Optimizer Algorithm

Tianhong Dai¹, Chunxue Sun¹, Jianping Huang¹, Qiancheng Xie¹, Shijie Cong¹, Xinwang Huang¹, and Kexin Li^2、*

Author Affiliations

¹College of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin 150040, Heilongjiang , China

²College of Artificial Intelligence, Wuxi Vocational College of Science and Technology, Wuxi 214000, Jiangsu , China

show less

DOI: 10.3788/LOP202259.1030003 Cite this Article Set citation alerts

Tianhong Dai, Chunxue Sun, Jianping Huang, Qiancheng Xie, Shijie Cong, Xinwang Huang, Kexin Li. Hyperspectral Wave Band Selection Based on Golden Sine and Chaotic Spotted Hyena Optimizer Algorithm[J]. Laser & Optoelectronics Progress, 2022, 59(10): 1030003 Copy Citation Text

show less

Fig. 1. Principle diagram of Golden-SA

Download full size

Fig. 2. Block diagram of hyperspectral band selection method

Download full size

Fig. 3. Flow chart of GSSHO algorithm

Download full size

Fig. 4. Salinas data set classification chart. (a) Original image; (b) GSSHO algorithm; (c) PSO algorithm; (d) SA algorithm; (e) GA algorithm; (f) GWO algorithm

Download full size

Fig. 5. Pavia Centre data set classification chart. (a) Original image; (b) GSSHO algorithm; (c) PSO algorithm; (d) SA algorithm; (e) GA algorithm; (f) GWO algorithm

Download full size

Algorithm	Parameter	Set value
GSSHO	Control factor $h$	［0，5］
	Number of candidate solutions $M$	［0.5，1］
	Individual $k$	2
PSO	Weight $ω$	［0.2，0.9］
	Learning factors $c_{1}, c_{2}$	1.49
	Speed $υ$	［0，6］
SA	Annealing coefficient $α_{T}$	0.99
SA	Initial temperature $T_{0}$	0.1
GA	Cross probability $C$	0.4
GA	Variation probability $M$	0.4
GWO	Synergy coefficient $a$	［0，2］

Table 1. Algorithm parameters involved in experiments

Band selection algorithm	Overall classification accuracy /%	Number of bands	Fitness function value	Kappa coefficient	Running time /s
GSSHO	93.32	32	0.0836	0.9283	10852
PSO	92.73	40	0.0974	0.9201	12360
SA	93.17	92	0.1372	0.9259	19863
GA	93.64	121	0.1619	0.9285	8979
GWO	93.08	94	0.1366	0.9241	10593

Table 2. Classification results of Salinas data set by each algorithm

Band selection algorithm	Overall classification accuracy /%	Number of bands	Fitness function value	Kappa coefficient	Running time /s
GSSHO	99.08	11	0.0315	0.9856	7038
PSO	98.57	15	0.0396	0.9787	7485
SA	99.02	37	0.0797	0.9849	8096
GA	99.14	58	0.1259	0.9868	4980
GWO	98.96	45	0.0932	0.9833	6975

Table 3. Classification results of Pavia Centre data set by each algorithm

Algorithm	GSSHO	PSO	SA	GA	GWO
Grapes_untrained	90.38%	89.06%	90.03%	90.55%	89.85%
Vinyard_untrained	68.72%	64.89%	68.28%	68.56%	67.56%

Table 4. Accuracy of each algorithm for two kinds of classification

Classification algorithm	Overall classification accuracy /%	Number of bands	Fitness function value
SVM	93.32	32	0.0836
KNN	90.45	43	0.1197
DT	87.91	41	0.1336
NB	77.07	35	0.2183

Table 5. Results of different classification algorithms in Salinas data set

Classification algorithm	Overall classification accuracy /%	Number of bands	Fitness function value
SVM	99.08	11	0.0315
KNN	98.18	39	0.0942
DT	97.19	21	0.0627
NB	88.71	15	0.1111

Table 6. Results of different classification algorithms in Pavia Centre data set

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information