Laser & Optoelectronics Progress, Volume. 59, Issue 4, 0428003(2022)
Band Selection Algorithm Based on Inter-Class Separability
Figures & Tables(8)
Fig. 3. SVM classification results of different algorithms. (a) Ground truth; (b) all band; (c) ABS algorithm; (d) proposed algorithm
Fig. 4. SVM classification results of different algorithms. (a) Ground truth; (b) all band; (c) ABS algorithm; (d) proposed algorithm
Table 1. Band selection results on PaviaU dataset
View tableTable 1. Band selection results on PaviaU dataset
Band IISF IIUF Band R 65 1078.3539 1089.0187 103 0.0388 67 1145.0336 1156.0180 101 0.0406 69 1213.6259 1225.0171 97 0.0409 71 1284.2393 1296.0161 99 0.0426 73 1356.8827 1369.0152 95 0.0461 75 1431.5357 1444.0158 93 0.0513 77 1508.1021 1521.0185 91 0.0555 79 1586.6852 1600.0215 89 0.0609 81 1667.2505 1681.0239 87 0.0700 83 1749.8526 1764.0246 85 0.0857
Table 2. Comparison of classification accuracy and time on PaviaU dataset
View tableTable 2. Comparison of classification accuracy and time on PaviaU dataset
Category All band ABS Proposed algorithm Asphalt /% 74.48 55.98 60.46 Meadow /% 74.78 70.87 76.23 Gravel /% 73.42 44.64 83.71 Tree /% 83.78 77.71 85.61 Painted metal sheet /% 99.55 98.74 99.03 Bare soil /% 67.93 69.85 88.59 Bitumen /% 87.97 91.65 95.56 Self-blocking-brick /% 78.52 80.53 66.24 Shadow /% 99.68 99.71 99.89 OA /% 76.57 70.54 77.88 Kappa /% 69.88 63.25 71.81 Time /s 3.81 1.53 1.63
Table 3. Band selection results on Salinas dataset
View tableTable 3. Band selection results on Salinas dataset
Band IISF IIUF Band R 126 1021.2952 1024.0004 86 0.1139 130 1085.9973 1089.0003 94 0.1373 134 1153.0147 1156.0003 98 0.1502 138 1222.4012 1225.0002 90 0.1750 142 1294.1586 1296.0002 82 0.2193 146 1368.2978 1369.0000 66 0.3250 150 1443.9285 1444.0000 102 0.3311 154 1520.8314 1521.0000 70 0.3649 158 1599.4054 1600.0004 78 0.3790 162 1680.2606 1681.0003 74 0.4038
Table 4. Comparison of classification accuracy and time on Salinas dataset
View tableTable 4. Comparison of classification accuracy and time on Salinas dataset
Category All band ABS Proposed algorithm Broccoli-green-weeds1 /% 95.62 99.30 95.97 Broccoli-green-weeds2 /% 99.70 90.47 96.19 Fallow /% 91.75 83.86 96.71 Fallow-rough-plow /% 97.63 99.57 99.14 Fallow-smooth /% 97.46 93.17 95.79 Stubble /% 98.91 99.01 98.66 Celery /% 99.44 99.22 90.42 Grapes-untrained /% 68.70 53.68 74.01 Soil-vinyard-develop /% 98.82 93.31 98.42 Corn-senesced-green-weed /% 88.01 71.84 82.15 Lettuce-romaine-4wk /% 86.42 97.19 90.64 Lettuce-romaine-5wk /% 99.53 99.38 98.29 Lettuce-romaine-6wk /% 97.82 97.60 97.38 Lettuce-romaine-7wk /% 92.06 96.45 90.37 Vinyard-untrained /% 58.31 72.17 54.36 Vineyard-vertical-trellis/ % 98.23 90.48 92.53 OA /% 85.70 82.04 85.00 Kappa /% 84.09 80.09 83.35 Time /s 15.22 3.85 3.68
Tools
Get Citation
Copy Citation Text
Liguo Zhang, Shengchun Sun, Lei Wang, Mei Jin, Yong Zhang, Bo Liu. Band Selection Algorithm Based on Inter-Class Separability[J]. Laser & Optoelectronics Progress, 2022, 59(4): 0428003
Paper Information
Category: Remote Sensing and Sensors
Received: Jan. 25, 2021
Accepted: Apr. 7, 2021
Published Online: Jan. 25, 2022
The Author Email: Mei Jin (506330828@qq.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20