Research Article

Local and Global Geometric Structure Preserving and Application to Hyperspectral Image Classification

Table 3

Comparison of class accuracy for Indian Pine scene (embedding , %).

ItemDimensionality reduction methodologies
Classifier Class ID PCA MFA LapLDA RP LPP LGSPP JGLDA LGGSP

5NN 1 64.86 50.00 45.95 70.59 44.44 66.67 36.3686.11
2 54.56 29.71 48.24 52.15 53.05 42.8055.3372.10
3 51.07 34.39 38.62 33.42 36.1446.08 29.8062.79
4 28.36 8.96 6.28 33.17 13.1121.57 11.4457.28
5 73.95 23.39 63.62 74.00 74.5976.73 55.8084.93
6 91.85 41.46 72.3895.50 86.6989.34 81.14 93.72
7 81.82 54.55 81.82 88.89100.0090.00 54.55100.00
8 92.82 58.54 92.64 85.71 88.3384.28 97.02 96.11
9100.00 0.00 60.0080.0075.0075.00 0.00100.00
10 67.58 16.17 36.69 59.79 51.8257.35 29.0972.99
11 67.96 37.33 60.59 69.19 62.8270.94 65.86 70.77
12 25.41 17.99 14.91 35.60 26.4334.87 28.8869.45
13 91.11 12.50 69.61 87.43 91.7691.38 87.8697.30
1494.28 53.31 72.98 91.12 86.5792.33 91.93 93.17
15 17.25 6.36 19.63 12.76 28.797.06 11.9554.14
16 79.17 31.58 30.67 73.97 78.0883.33 86.49 84.93

RBF-SVM 178.38 0.00 29.7379.41 52.7869.70 75.76 61.11
2 67.82 0.00 49.5458.53 58.9349.62 56.2474.24
363.81 0.00 37.2646.13 54.1440.77 48.21 62.01
4 49.25 0.00 3.8650.50 52.4335.29 25.8762.14
584.11 0.00 56.2583.41 78.0974.50 72.10 84.02
693.16 0.00 67.9591.30 85.8091.74 80.99 90.88
7 90.91 0.00 72.7355.56 100.0080.00 81.82100.00
8 96.30 1.55 94.5488.89 93.3684.97 89.6896.80
9100.00 0.00 40.00100.00 100.0075.00 0.00 80.00
10 57.45 0.00 29.7861.16 57.8662.74 44.0967.54
11 74.16100.00 72.3776.88 69.4777.87 34.49 82.49
12 57.12 0.00 15.7958.72 41.5938.75 41.1380.36
13 91.11 0.00 69.6193.19 93.9688.51 90.75 85.95
14 92.92 0.00 78.0393.25 88.7589.71 76.3793.34
15 36.84 0.00 35.5843.92 45.1528.24 43.4450.00
1686.11 0.00 48.0079.45 73.9780.77 78.38 60.27