Hyperspectral image classification based on multi-scale hybrid convolutional network

类别	类名	中文类名	样本数
1	Alfalfa	苜蓿	46
2	Corn-notill	免耕玉米	1 428
3	Corn-mintill	少耕玉米	830
4	Corn	玉米	237
5	Grass-pasture	牧草	483
6	Grass-trees	草地-树木	730
7	Grass-pasture-mowed	收割过的草-牧场	28
8	Hay-windrowed	干草料堆	478
9	Oats	燕麦	20
10	Soybean-notill	免耕大豆	972
11	Soybean-mintill	少耕大豆	2 455
12	Soybean-clean	修剪后的大豆	593
13	Wheat	小麦	205
14	Woods	树木	1 265
15	Buildings-Grass-Trees-Drivers	建筑物-草-树-车道	386
16	Stone-Steel-Towers	石头-钢材-塔楼	93
Total			10 249

Table 2. Sample distribution of PaviaU dataset

View table

View in Article

Table 2. Sample distribution of PaviaU dataset

类别	类名	中文类名	样本数
1	Asphalt	柏油马路	6 631
2	Meadows	草地	18 649
3	Gravel	砂砾	2 099
4	Trees	树木	3 064
5	Painted metal sheets	金属板	1 345
6	Bare Soil	裸露土壤	5 029
7	Bitumen	沥青屋顶	1 330
8	Self-Blocking Bricks	砖路	3 682
9	Shadows	阴影	947
Total			42 776

Table 3. Influence of B value on classification accuracy of IP dataset
View table
View in Article
Table 3. Influence of B value on classification accuracy of IP dataset
B OA B OA
10 0.984 4 25 0.987 9
15 0.982 7 30 0.986 1
20 0.985 5 35 0.982 4

Table 4. Influence of B value on classification accuracy of PaviaU dataset
View table
View in Article
Table 4. Influence of B value on classification accuracy of PaviaU dataset
B OA B OA
10 0.993 6 25 0.999 0
15 0.997 3 30 0.997 8
20 0.996 5 35 0.995 7

Table 5. Influence of S on classification accuracy of IP dataset
View table
View in Article
Table 5. Influence of S on classification accuracy of IP dataset
B OA B OA
10 0.983 7 25 0.981 4
15 0.982 8 30 0.987 9
20 0.984 0 35 0.985 8

Table 6. Influence of S on classification accuracy of PaviaU dataset
View table
View in Article
Table 6. Influence of S on classification accuracy of PaviaU dataset
B OA B OA
10 0.995 7 25 0.997 9
15 0.999 0 30 0.995 8
20 0.995 1 35 0.997 4

Table 7. Classification results of different algorithms in IP dataset

View table

View in Article

Table 7. Classification results of different algorithms in IP dataset

类别	SVM	3D-CNN	HybridSN	MS-3D	MS-Hybrid-A
1	0.595 2	0.634 1	0.926 8	1.000 0	0.951 2
2	0.766 7	0.794 6	0.737 7	0.933 1	0.971 2
3	0.637 2	0.714 9	0.990 6	0.990 6	0.988 0
4	0.621 5	0.521 1	0.859 2	1.000 0	1.000 0
5	0.917 2	0.887 4	0.979 3	0.979 3	0.990 8
6	0.958 9	0.969 6	0.993 9	0.977 2	0.987 8
7	0.653 8	0.840 0	0.520 0	1.000 0	1.000 0
8	0.972 2	0.990 7	0.990 7	1.000 0	1.000 0
9	0.444 4	0.611 1	0.666 7	0.944 4	1.000 0
10	0.682 3	0.795 4	0.971 4	0.971 4	0.993 1
11	0.857 5	0.855 2	0.931 2	0.991 9	0.991 4
12	0.702 2	0.706 0	0.971 9	0.975 7	0.973 8
13	0.983 8	0.945 9	0.897 3	0.973 0	1.000 0
14	0.956 1	0.963 1	1.000 0	0.995 6	0.998 2
15	0.497 1	0.772 3	0.968 3	0.976 9	0.994 2
16	0.881 0	0.940 5	0.881 0	0.916 7	0.892 8
OA	0.810 1	0.840 7	0.930 2	0.978 4	0.987 9
AA	0.758 0	0.808 9	0.892 9	0.976 6	0.983 3
Kappa	0.782 5	0.817 8	0.920 6	0.975 4	0.986 2

Table 8. Classification results of different algorithms in PaviaU dataset

View table

View in Article

Table 8. Classification results of different algorithms in PaviaU dataset

类别	SVM	3D-CNN	HybridSN	MS-3D	MS-Hybrid-A
1	0.936 0	0.988 1	0.989 9	0.996 6	1.000 0
2	0.984 6	0.994 5	0.999 8	0.998 2	1.000 0
3	0.794 1	0.896 2	0.944 4	0.985 7	0.987 8
4	0.946 7	0.977 2	0.986 9	0.993 8	0.998 5
5	0.990 1	0.990 9	0.995 0	1.000 0	1.0000
6	0.863 7	0.988 5	0.999 8	1.000 0	1.000 0
7	0.847 1	0.960 7	0.999 2	1.000 0	1.000 0
8	0.906 5	0.896 5	0.955 3	0.979 5	0.999 7
9	1.000 0	0.929 6	0.931 9	0.993 0	0.985 9
OA	0.940 3	0.975 7	0.989 1	0.995 6	0.999 0
AA	0.918 8	0.958 0	0.978 0	0.994 1	0.996 9
Kappa	0.920 5	0.967 8	0.985 6	0.994 2	0.998 6

Tools

Get Citation

Copy Citation Text

Yun YANG, Yao ZHOU, Jia-ning CHEN. Hyperspectral image classification based on multi-scale hybrid convolutional network[J]. Chinese Journal of Liquid Crystals and Displays, 2023, 38(3): 368

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Research Articles

Received: Jul. 1, 2022

Accepted: --

Published Online: Apr. 3, 2023

The Author Email: Yun YANG (yangyun0806@163.com)

DOI:10.37188/CJLCD.2022-0225

Topics

Table 1. Sample distribution of IP dataset

Table 1. Sample distribution of IP dataset

Table 2. Sample distribution of PaviaU dataset

Table 2. Sample distribution of PaviaU dataset

Table 3. Influence of B value on classification accuracy of IP dataset

Table 3. Influence of B value on classification accuracy of IP dataset

Table 4. Influence of B value on classification accuracy of PaviaU dataset

Table 4. Influence of B value on classification accuracy of PaviaU dataset

Table 5. Influence of S on classification accuracy of IP dataset

Table 5. Influence of S on classification accuracy of IP dataset

Table 6. Influence of S on classification accuracy of PaviaU dataset

Table 6. Influence of S on classification accuracy of PaviaU dataset

Table 7. Classification results of different algorithms in IP dataset

Table 7. Classification results of different algorithms in IP dataset

Table 8. Classification results of different algorithms in PaviaU dataset

Table 8. Classification results of different algorithms in PaviaU dataset