Hyperspectral Image Classification Algorithm Based on Space-Spectral Weighted Nearest Neighbor

Lei Ji; Xin Zhang; Limei Zhang; Zhang Wen

doi:10.3788/LOP57.061013

Journals >Laser & Optoelectronics Progress >Volume 57 >Issue 6 >Page 061013 > Article

Laser & Optoelectronics Progress
Vol. 57, Issue 6, 061013 (2020)

Hyperspectral Image Classification Algorithm Based on Space-Spectral Weighted Nearest Neighbor

Lei Ji¹, Xin Zhang^1,*, Limei Zhang², and Zhang Wen¹

Author Affiliations

¹College of Big Data and Information Engineering, Guizhou University, Guiyang Guizhou 550025, China

²College of Optoelectronic Engineering, Chongqing University, Chongqing 400044, China

show less

DOI: 10.3788/LOP57.061013 Cite this Article Set citation alerts

Lei Ji, Xin Zhang, Limei Zhang, Zhang Wen. Hyperspectral Image Classification Algorithm Based on Space-Spectral Weighted Nearest Neighbor[J]. Laser & Optoelectronics Progress, 2020, 57(6): 061013 Copy Citation Text

show less

Process of removing background point. (a) Original image; (b) random sample points; (c) non-nearest neighbor sample points; (d) processing non-nearest neighbor sample points; (e) filtered sample points

Fig. 1. Process of removing background point. (a) Original image; (b) random sample points; (c) non-nearest neighbor sample points; (d) processing non-nearest neighbor sample points; (e) filtered sample points

Download full size

Fig. 2. Indian Pines dataset. (a) False-color image; (b) ground-type survey map;(c) spectral curves^[14]

Download full size

Fig. 3. PaviaU dataset. (a) False-color image; (b) ground-type survey map; (c) spectral curves^[14]

Download full size

Fig. 4. OA of Indian Pines dataset with different spatial windows

Download full size

Fig. 5. OA of different algorithms with different percentages of training samples

Download full size

Fig. 6. Classification results of different algorithms in Indian Pines dataset. (a) NN; (b) SRC; (c) SVM; (d) WSSD-KNN; (e)SSNN; (f) SSWNN

Download full size

Fig. 7. OA of PaviaU dataset with different spatial windows

Download full size

Fig. 8. OA of different algorithms with different percentages of training samples

Download full size

Fig. 9. Classification results of different algorithms in PaviaU dataset. (a) NN; (b) SRC; (c) SVM; (d) WSSD-KNN; (e) SSNN; (f) SSWNN

Download full size

Grade	Category	Trainingsample set	Testsample set	Classification accuracy /%
Grade	Category	Trainingsample set	Testsample set	NN	SRC	SVM	WSSD-KNN	SSNN	SSWNN
1	Alfalfa	10	36	48.28	55.56	69.44	92.11	97.22	100.00
2	Corn-notill	143	1285	58.86	55.98	72.13	89.13	94.17	98.24
3	Corn-min	83	747	51.47	54.03	69.23	84.14	93.84	94.55
4	Corn	24	213	44.15	41.40	57.61	82.76	88.00	99.00
5	Grass/Pasture	48	435	85.30	82.37	86.71	97.03	100.00	98.82
6	Grass/Tress	73	657	84.30	79.92	90.61	98.31	93.94	96.98
7	Grass Pasture mowed	10	18	50.00	60.61	69.23	69.23	76.67	80.77
8	Hay-windrowed	48	430	91.29	93.10	96.52	100.00	98.62	100.00
9	Oats	10	10	16.67	14.29	38.46	62.50	65.22	63.16
10	Soybean-notill	97	875	59.55	58.39	73.99	87.76	95.27	95.74
11	Soybean-min	246	2209	69.40	69.54	81.82	92.36	95.29	96.02
12	Soybean-clean	59	534	48.35	46.72	79.22	82.99	91.62	96.17
13	Wheat	21	184	85.43	86.44	94.27	98.92	94.59	94.38
14	Woods	127	1138	90.54	89.05	93.80	97.82	99.12	97.09
15	Buildings-Grass-Tree-Drives	39	347	40.94	51.44	63.95	91.88	98.18	99.37
16	Stone-steel-towers	10	83	96.34	98.68	98.65	98.80	93.24	94.74
OA				68.70	68.71	80.66	91.74	95.31	96.75
AA				63.80	64.84	77.23	89.11	92.19	94.06
Kappa				0.643	0.642	0.780	0.906	0.947	0.963

Table 1. Classification accuracy of different classes in Indian Pines dataset for different algorithms

Grade	Category	Trainingsample set	Testsample set	Classification accuracy /%
Grade	Category	Trainingsample set	Testsample set	NN	SRC	SVM	WSSD-KNN	SSNN	SSWNN
1	Asphalt	398	6233	91.61	93.13	93.09	97.91	100.00	99.64
2	Meadows	1119	17530	87.79	87.09	93.23	97.73	99.90	99.49
3	Gravel	126	1973	65.98	65.74	84.46	96.51	61.76	97.11
4	Trees	184	2880	94.34	94.97	95.35	99.57	98.77	97.82
5	Sheets	81	1264	99.42	99.75	99.26	99.59	99.91	96.52
6	Soil	302	4727	71.82	71.93	87.65	96.47	99.78	99.49
7	Bitumen	80	1250	69.08	68.04	87.57	91.46	77.68	98.18
8	Bricks	221	3461	65.22	66.95	79.19	93.68	96.08	93.59
9	Shadows	57	890	99.75	99.88	99.40	99.65	96.13	96.34
OA				83.83	83.91	91.24	97.21	95.57	98.54
AA				82.78	83.05	91.02	96.95	92.22	97.57
Kappa				0.783	0.784	0.883	0.963	0.941	0.981

Table 2. Classification accuracy of different classes in PaviaU dataset for different algorithms

Lei Ji, Xin Zhang, Limei Zhang, Zhang Wen. Hyperspectral Image Classification Algorithm Based on Space-Spectral Weighted Nearest Neighbor[J]. Laser & Optoelectronics Progress, 2020, 57(6): 061013

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information

微信扫一扫：分享

微信扫一扫：分享