PFWG改进的CNN多光谱遥感图像分类 下载: 1188次
PFWG Improved CNN Multispectra Remote Sensing Image Classification
西安建筑科技大学信息与控制工程学院, 陕西 西安 710055
图 & 表
图 1. CNN示意图
Fig. 1. Schematic of CNN
下载图片 查看原文
图 2. 研究区影像。(a)研究区1;(b)研究区2
Fig. 2. Research area image. (a) Study area 1; (b) study area 2
下载图片 查看原文
图 3. 跨轨道光照辐射校正示意图
Fig. 3. Cross-track illumination radiation correction diagram
下载图片 查看原文
图 4. 辐射校正示意图
Fig. 4. Radiation correction diagram
下载图片 查看原文
图 5. MNF结果图
Fig. 5. MNF result graph
下载图片 查看原文
图 6. MNF示意图
Fig. 6. Diagram of MNF
下载图片 查看原文
图 7. BP神经网络分类结果图
Fig. 7. BP neural network classification result graph
下载图片 查看原文
图 8. CNN分类结果图
Fig. 8. CNN classification result graph
下载图片 查看原文
图 9. PFWG改进CNN分类结果图
Fig. 9. PFWG improved CNN classification result graph
下载图片 查看原文
图 10. 分类结果对比图。 (a) BP神经网络分类结果;(b) CNN分类结果;(c) PFWG改进的CNN分类结果
Fig. 10. Classification result comparison graph. (a) BP neural network classification result; (b) CNN classification result; (c) PFWG improved CNN classification result
下载图片 查看原文
表 1实验区土地覆盖类型及特征解译
Table1. Land cover classification system and interpretation marks in the study area
Image | Feature type | Image characteristic |
---|
| Building | Buildings are arranged neatly and there are roads passing through.The colors are gray and dark gray. | | Water | Uniform texture, smooth borders, blue-green or dark blue. | | Green area | Blocky distribution, the color characteristics are more obvious, dark green, light blue. | | Nudation | Texture shows a pattern, there are roads through, the color is purple, gray purple. |
|
查看原文
表 2BP神经网络分类结果混淆矩阵
Table2. BP neural network classification result confusion matrix%
Classification | Building | Nudation | Water | Greenbelt | Aggregate |
---|
Building | 91.44 | 19.73 | 6.07 | 3.47 | 30.1 | Nudation | 8.56 | 80.27 | 0 | 0 | 22.1 | Water | 0 | 0 | 93.93 | 0 | 23.3 | Greenbelt | 0 | 0 | 0 | 96.53 | 24.5 | Aggregate | 100 | 100 | 100 | 100 | 100 |
|
查看原文
表 3CNN分类结果混淆矩阵
Table3. CNN classification result confusion matrix%
Classification | Building | Nudation | Water | Greenbelt | Aggregate |
---|
Building | 93.12 | 17.85 | 2.94 | 0.71 | 28.6 | Nudation | 6.88 | 82.15 | 0 | 0 | 22.2 | Water | 0 | 0 | 97.06 | 0 | 24.2 | Greenbelt | 0 | 0 | 0 | 99.29 | 25.0 | Aggregate | 100 | 100 | 100 | 100 | 100 |
|
查看原文
表 4PFWG改进CNN分类结果混淆矩阵
Table4. PFWG improved CNN classification result confusion matrix%
Classification | Building | Nudation | Water | Greenbelt | Aggregate |
---|
Building | 96.63 | 15.48 | 0.82 | 0 | 28.3 | Nudation | 3.37 | 84.52 | 0 | 0 | 21.9 | Water | 0 | 0 | 99.18 | 0 | 24.8 | Greenbelt | 0 | 0 | 0 | 100 | 25.0 | Aggregate | 100 | 100 | 100 | 100 | 100 |
|
查看原文
表 5分类精度评价矩阵
Table5. Classification accuracy evaluation matrix
Parameter | BP neural network | Convolutional neural network | Improved algorithm |
---|
Kappa coefficient | 0.88 | 0.90 | 0.94 | Classification speed /min | 25 | 7 | 4 | Overall accuracy /% | 91.63 | 92.82 | 93.73 |
|
查看原文
王民, 樊潭飞, 贠卫国, 王稚慧. PFWG改进的CNN多光谱遥感图像分类[J]. 激光与光电子学进展, 2019, 56(3): 031003. Min Wang, Tanfei Fan, Weiguo Yun, Zhihui Wang. PFWG Improved CNN Multispectra Remote Sensing Image Classification[J]. Laser & Optoelectronics Progress, 2019, 56(3): 031003.