Feature Extraction of Hyperspectral Images Based on Semi-Supervised Locality Preserving Projection with Spatial-Correlation

Fig. 4. Overall classification accuracy of different labels in Indian Pines dataset with different algorithms. (a) 10 labels; (b) 20 labels; (c) 30 labels

Download full size

Fig. 5. Overall classification accuracy of different labels in Pavia University dataset with different algorithms. (a) 10 labels; (b) 20 labels; (c) 30 labels

Download full size

Table 1. Highest overall classification accuracy of different labels in Indian Pines dataset with different algorithms (average±standard variance /%)

View table

Table 1. Highest overall classification accuracy of different labels in Indian Pines dataset with different algorithms (average±standard variance /%)

Algorithm	Highest overall classification accuracy
	10 labels		20 labels		30 labels
	1-NN	SVM	1-NN	SVM	1-NN	SVM
PCA	72.81±0.59	75.02±1.43	74.03±0.27	75.94±0.82	75.04±0.51	76.59±1.38
LPP	73.29±0.20	76.96±0.78	75.76±0.14	77.51±1.05	76.17±0.44	78.89±0.67
LSDA	74.69±2.67	78.20±3.69	76.61±1.92	80.92±1.90	77.96±2.84	81.15±2.17
NPE	74.21±2.20	73.21±2.69	75.84±1.25	75.16±1.27	77.06±2.16	78.45±1.75
SSLPP	75.89±1.54	77.25±1.54	77.70±1.47	81.38±1.59	78.68±1.95	82.09±0.70
SSDR-PCP	78.59±0.50	79.67±0.62	79.89±0.39	80.89±0.73	80.30±0.63	82.46±0.78
LPP-SCSSFE	82.48±0.74	84.62±0.78	83.40±0.42	86.48±0.67	83.95±0.59	87.50±0.37

Table 2. Classification accuracies of different types of samples in Indian Pines dataset with different algorithms

View table

Table 2. Classification accuracies of different types of samples in Indian Pines dataset with different algorithms

Class	Classification accuracy /%
Class	PCA	LPP	LSDA	NPE	SSLPP	SSDR-PCP	LPP-SCSSFE
1	68.17	63.88	64.84	60.93	75.39	76.94	73.46
2	63.00	61.38	65.63	66.25	62.88	62.25	66.75
3	85.07	83.40	64.46	80.57	84.26	85.35	91.39
4	89.29	88.29	90.86	88.86	93.00	93.86	94.86
5	97.11	96.33	95.42	97.75	100.00	98.88	100.00
6	66.91	67.09	68.97	68.20	72.82	74.76	70.48
7	74.72	74.19	77.07	77.37	77.74	79.57	81.28
8	52.66	53.11	67.00	62.27	56.09	55.12	68.74
9	90.31	90.96	94.37	95.95	90.65	91.01	97.67
10	52.58	51.12	65.96	60.28	62.51	61.12	68.54
OA	75.04	76.17	77.96	77.06	78.68	80.30	83.95
AA	73.98	72.98	75.46	75.84	77.53	77.89	81.32
κ	72.38	72.65	72.50	75.35	74.22	76.70	80.99

Table 3. Highest overall classification accuracy of different labels in Pavia University dataset with different algorithms (average±standard variance /%)

View table

Table 3. Highest overall classification accuracy of different labels in Pavia University dataset with different algorithms (average±standard variance /%)

Algorithm	Highest overall classification accuracy
	10 labels		20 labels		30 labels
	1-NN	SVM	1-NN	SVM	1-NN	SVM
PCA	75.36±0.09	75.83±1.49	76.03±0.43	75.32±1.55	76.62±0.17	76.37±2.99
LPP	78.19±1.08	78.22±0.48	78.36±0.31	78.51±1.17	79.40±0.99	79.47±0.48
LSDA	79.90±2.55	81.36±2.97	81.07±2.48	83.72±2.42	81.85±2.14	84.44±2.15
NPE	78.96±2.03	78.92±2.08	79.18±2.43	79.37±2.82	79.92±1.72	81.00±3.05
SSLPP	80.68±0.42	81.22±0.42	82.26±0.43	83.67±0.60	82.57±0.40	84.29±0.23
SSDR-PCP	83.96±0.39	83.18±0.84	84.99±0.31	84.23±0.96	86.61±0.12	85.38±1.28
LPP-SCSSFE	87.84±0.72	88.11±0.47	89.10±0.65	90.04±0.12	89.57±0.40	91.29±0.23

Table 4. Classification accuracies of different types of samples in Pavia University dataset with different algorithms

View table

Table 4. Classification accuracies of different types of samples in Pavia University dataset with different algorithms

Class	Classification accuracy /%
Class	PCA	LPP	LSDA	NPE	SSLPP	SSDR-PCP	LPP-SCSSFE
1	81.91	85.34	75.05	74.20	83.64	85.67	86.47
2	92.61	93.64	89.92	86.17	95.79	95.74	96.65
3	71.44	70.61	65.20	64.62	72.93	72.67	82.34
4	84.28	86.26	94.03	94.13	91.83	91.29	93.38
5	99.54	99.35	99.70	99.39	99.77	99.54	99.85
6	59.99	69.15	65.71	64.41	72.35	72.31	78.44
7	86.57	85.31	85.77	80.46	86.54	85.00	90.75
8	70.31	72.68	68.37	73.63	76.97	72.92	78.77
9	98.00	98.00	98.89	98.67	97.24	97.38	98.91
OA	76.62	79.40	81.85	79.92	82.57	86.61	89.57
AA	74.47	76.08	74.26	73.57	77.71	77.25	80.56
κ	73.73	76.63	78.48	76.61	78.96	83.01	86.70

Tools

Get Citation

Copy Citation Text

Dongmei Huang, Xiaotong Zhang, Minghua Zhang, Wei Song, Yan Wang. Feature Extraction of Hyperspectral Images Based on Semi-Supervised Locality Preserving Projection with Spatial-Correlation[J]. Laser & Optoelectronics Progress, 2019, 56(2): 021003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image Processing

Received: May. 21, 2018

Accepted: Jul. 30, 2018

Published Online: Aug. 1, 2019

The Author Email: Zhang Minghua (mhzhang@shou.edu.cn)

DOI:10.3788/LOP56.021003

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology