Laser & Optoelectronics Progress, Volume. 60, Issue 12, 1210015(2023)
Band Selection of Hyperspectral Images Based on Fuzzy C-Means Clustering and Firefly Algorithm
Figures & Tables(10)
Fig. 2. Classification accuracy of different methods on Indian Pines dataset. (a) OA of the SVM classifier; (b) OA of the KNN classifier; (c) Kappa coefficient of the SVM classifier; (d) Kappa coefficient of the KNN classifier
Fig. 3. Classification accuracy of different methods on PaviaU dataset. (a) OA of SVM classifier; (b) OA of KNN classifier; (c) Kappa coefficient of SVM classifier; (d) Kappa coefficient of KNN classifier
Fig. 4. Classification accuracy of different methods in 28 selected bands on Indian Pines dataset using different proportions of training samples. (a) OA of SVM classifier; (b) OA of KNN classifier; (c) Kappa coefficient of SVM classifier; (d) Kappa coefficient of KNN classifier
Fig. 5. Classification accuracy of different methods in 18 selected bands on PaviaU dataset using different proportions of training samples. (a) OA of SVM classifier; (b) OA of KNN classifier; (c) Kappa coefficient of SVM classifier; (d) Kappa coefficient of KNN classifier
Fig. 6. Spectral characteristic curves of land type on Indian Pines dataset. (a)-(p) Characteristic curves of land type 1-16, respectively
Fig. 7. Spectral characteristic curves of land type on PaviaU dataset. (a)~(i) Characteristic curves of land type 1-9, respectively
Table 1. Calculation time of different methods
View tableTable 1. Calculation time of different methods
Dataset ISSC EGCSR-R EGCSR-C OPBS SpaBS FCM FCM-FA Indian Pines 9.47 0.12 8.37 4.23 127.13 42.01 163.69 PaviaU 40.63 0.54 2.19 11.22 542.49 258.19 553.92
Table 2. Calculation time of FCM-FA for different fireflies
View tableTable 2. Calculation time of FCM-FA for different fireflies
Dataset N=5 N=8 N=10 N=12 N=15 Indian Pines 155.19 160.91 163.69 170.61 176.33 PaviaU 523.84 539.57 553.92 563.52 572.99
Table 3. Effect of the setting of number of fireflies on results of FCM-FA band selection (band quality is indirectly evaluated through classification accuracy)
View tableTable 3. Effect of the setting of number of fireflies on results of FCM-FA band selection (band quality is indirectly evaluated through classification accuracy)
Dataset Classifier Parameter N=5 N=8 N=10 N=12 N=15 Indian Pines SVM OA 80.368 83.340 83.722 83.568 85.134 AA 80.216 83.334 83.714 83.536 85.035 Kappa 77.514 80.929 81.366 81.193 83.008 KNN OA 66.557 72.434 72.655 72.686 74.999 AA 66.079 72.237 72.521 72.603 74.886 Kappa 61.708 68.491 68.743 68.790 71.400 PaviaU SVM OA 94.214 94.214 94.214 94.214 94.214 AA 94.199 94.199 94.199 94.199 94.199 Kappa 92.315 92.315 92.315 92.315 92.315 KNN OA 89.379 89.379 89.379 89.379 89.379 AA 89.393 89.393 89.393 89.393 89.393 Kappa 85.720 85.720 85.720 85.720 85.720
Tools
Get Citation
Copy Citation Text
Zhou Zhang, Xu Sun, Rong Liu, Lianru Gao. Band Selection of Hyperspectral Images Based on Fuzzy C-Means Clustering and Firefly Algorithm[J]. Laser & Optoelectronics Progress, 2023, 60(12): 1210015
Paper Information
Category: Image Processing
Received: Mar. 28, 2022
Accepted: Jun. 22, 2022
Published Online: Jun. 5, 2023
The Author Email: Xu Sun (sunxu@aircas.ac.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20