step interval of 5. Using cross-validation, the regularization parameters of the ISSC algorithm on the Indian Pines and PaviaU data sets are set to 0.1 and 0.05, respectively. In Figure 3(a), the ISSC algorithm outperforms other methods and obtains a higher OCA compared to using all the bands when the band subset size is over 15. The MVPCA behaves worse than other methods. ISSC and LP achieve the best OCA results, as shown in Figure 3(b). The FDPC and MVPCA methods perform worse than other methods. Moreover, Tables 4 and 5 and Figures 1 and 2 present the detailed classification information and classification maps. The band subset sizes of both data sets are manually set to be 30. The observation coincides with that of Figure 3 and supports the performance description of all the methods. In addition, Tables 6 and 7 list the computing time of six different methods on the two data sets. All methods are implemented in MATLAB 2014a and run on a Windows 7 computer with an Intel i5-4570 Quad Core processor and 8 GB of random-access memory. Every method takes more time when the number of selected 85 95 80 90 OCA (%) OCA (%) 75 70 MVPCA MMCA WaluDI LP FDPC ISSC All Bands 65 60 55 50 5 10 15 20 25 30 35 40 45 50 The Number of Selected Bands (a) 55 MVPCA MMCA WaluDI LP FDPC ISSC All Bands 85 60 80 5 10 15 20 25 30 35 40 45 50 The Number of Selected Bands (b) 55 60 FIGURE 3. Overall classification accuracy curves of different band selection methods for the (a) Indian Pines and (b) PaviaU data sets. TABLE 4. THE CLASSIFICATION ACCURACIES (%) ACHIEVED USING 30 SELECTED BANDS FROM THE INDIAN PINES DATA SET. MVPCA MMCA WaluDI LP FDPC ISSC ALL BANDS Alfalfa 65.85 63.14 51.22 29.27 36.59 34.15 36.59 No-till corn 58.52 63.66 69.11 61.25 69.03 75.8 75.41 Minimal-till corn 41.5 62.92 57.7 46.72 66.27 65.19 66.8 Corn 47.89 59.15 55.4 48.83 56.81 53.05 59.14 Grass/pasture 81.38 89.2 86.44 87.36 82.99 82.07 82.53 Grass/trees 92.24 96.04 93.61 91.32 93.3 96.65 96.04 Mowed grass/pasture 44 76 44 68 32 92 56 Windrowed hay 98.6 89.53 96.28 98.14 94.88 96.74 98.61 Oats 27.78 50 22.22 33.33 22.22 61.11 38.89 No-till soybeans 59.09 68.69 75.89 63.2 66.29 79.43 66.4 Minimal-till soybean 76.14 75.55 78.27 79.95 81.62 82.53 80.76 Clean soybean 41.76 69.48 67.6 56.93 70.22 83.33 69.85 Wheat 92.93 95.65 96.74 96.2 96.74 96.2 98.91 Woods 93.59 95.52 94.38 93.59 93.5 96.84 94.38 Buildings/grass/trees/drives 41.79 51.01 49.28 51.87 48.41 53.6 52.74 Stone/steel towers 92.86 42.86 71.43 66.67 78.57 80.95 89.29 ACA 65.99 71.79 69.35 67.04 68.09 76.85 72.65 OCA 70.18 75.79 77.15 73.5 77.49 81.61 79.12 KC 65.83 72.27 73.86 69.56 74.22 78.98 76.05 NOTE: ACA: average classification accuracy; KC: Kappa coefficient. 132 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE JUNE 2019