Remote Sensing Technology and Application, Volume. 40, Issue 4, 1002(2025)
Optimal Feature Selection for Forest Disturbance Monitoring
Figures & Tables(21)
Fig. 4. JM distance of different features for forest fire monitoring （in descending order）
Fig. 5. JM distance of different features for deforestation monitoring （in descending order）
Fig. 6. JM distance of different forest geological disaster monitoring features （in descending order）
Fig. 7. The relationship between the number of features and recall of forest fire
Fig. 8. The relationship between the number of features and recall of deforestation
Fig. 9. Relationship between the number of feature and recall of forest geological disaster
Fig. 11. The quartile value range diagram of the reflection features of three types of forest disturbance
Fig. 12. JM distance heat map of texture features of three types of forest disturbance
Fig. 13. JM distance radar map of index features of three types of forest disturbance
Fig. 14. 3D scatter plots of optimal features of the three types of forest disturbance
Fig. 15. Classification results of three types of forest disturbance without feature compression and with feature compression
Table 1.
Data details View tableView in ArticleTable 1.
Data details 扰动类型 数据类型 扰动后影像
获取时间
扰动前影像
获取时间
森林火灾 Sentinel-1 SAR GRD 2020-04-02 2020-03-09 Sentinel-2 MSI Level-2A 2020-04-09 2020-03-05 森林砍伐 Sentinel-1 SAR GRD 2022-09-04 2022-08-11 Sentinel-2 MSI Level-2A 2022-09-05 2022-08-11 森林地质灾害 Sentinel-1 SAR GRD 2023-05-05 2023-04-11 Sentinel-2 MSI Level-2A 2023-05-07 2023-04-07
Table 2.
Candidate features View tableView in ArticleTable 2.
Candidate features 特征类型 特征指标 反射特征 B1、B2、B3、B4、B5、B6、B7、B8、B8A、B9、B11、B12 散射特征 VV、VH、q、mRVI、PRVI、RFDI 纹理特征 Mean、Variance、Homogeneity、Contrast、Dissimilarity、Entropy、Second Moment、Correlation 指数特征 NDVI 、NBR、SR 、DVI、PBI
Table 3.
Ranking of importance and accuracy evaluation of forest fire features View tableView in ArticleTable 3.
Ranking of importance and accuracy evaluation of forest fire features 变量个数 新增变量 JM 距离 准确率
/%
召回率
/%
精确度
/%
F1 score 1 B9 1.573 75.40 53.05 97.83 0.688 2 Mean-B9 1.537 91.87 88.41 95.34 0.917 3 Mean-B8A 1.465 92.17 88.80 95.56 0.921 4 B8A 1.457 92.25 89.00 95.57 0.922 5 Mean-B7 1.440 93.35 90.96 95.86 0.933 6 Mean-B8 1.435 93.35 91.55 95.30 0.934 7 B7 1.430 93.35 91.55 95.30 0.934 8 Mean-B6 1.426 93.35 91.36 95.48 0.934 9 B6 1.414 93.36 91.36 95.48 0.934 10 B8 1.337 93.35 91.36 95.48 0.934 11 DVI 1.093 93.35 91.36 95.48 0.934 12 Contrast-B2 1.041 93.96 91.36 96.67 0.939 13 NBR 1.034 93.86 91.16 96.67 0.938
Table 4.
Importance ranking and accuracy evaluation of deforestation features View tableView in ArticleTable 4.
Importance ranking and accuracy evaluation of deforestation features 变量个数 新增变量 JM距离 准确率
/%
召回率
/%
精确度
/%
F1 score 1 B9 1.961 83.51 67.65 97.87 0.800 2 Mean-B9 1.954 92.83 87.50 97.54 0.922 3 Mean-B7 1.921 90.68 83.09 97.14 0.897 4 B7 1.916 90.68 83.09 97.14 0.897 5 Mean-B6 1.904 91.76 85.29 97.48 0.910 6 Mean-B8A 1.899 87.81 77.21 97.22 0.861 7 B6 1.895 87.81 77.21 97.22 0.861 8 B8A 1.889 87.81 77.21 97.22 0.861 9 Mean-B8 1.880 86.38 74.26 97.12 0.842 10 NBR 1.851 86.38 74.26 97.12 0.842 11 DVI 1.808 86.38 74.26 97.12 0.842 12 NDVI 1.801 86.38 74.26 97.12 0.842 13 B8 1.741 86.38 74.26 97.12 0.842 14 PBI 1.574 86.02 73.53 97.09 0.837 15 SR 1.573 85.66 72.79 97.06 0.832
Table 5.
Ranking of importance and accuracy evaluation of forest geological disaster features View tableView in ArticleTable 5.
Ranking of importance and accuracy evaluation of forest geological disaster features 变量个数 新增变量 JM距离 准确率
/%
召回率
/%
精确度
/%
F1 score 1 NDVI 1.999 92.44 76.47 99.41 0.864 2 NBR 1.995 94.15 81.90 99.45 0.898 3 DVI 1.974 94.29 82.35 99.45 0.901 4 PBI 1.968 93.01 78.28 99.43 0.876 5 SR 1.917 93.30 79.19 99.43 0.882 6 Mean-B8A 1.892 94.44 83.71 98.40 0.905 7 B8A 1.891 94.44 83.71 98.40 0.905 8 Mean-B7 1.842 94.29 82.35 99.45 0.901 9 Mean-B8 1.841 92.72 77.83 98.85 0.871 10 B7 1.841 92.72 77.83 98.85 0.871 11 B9 1.797 93.01 78.28 99.43 0.876 12 Mean-B4 1.789 85.16 53.39 99.16 0.694 13 Mean-B9 1.780 84.17 49.77 100.00 0.666 14 Mean-B6 1.775 82.88 45.70 100.00 0.627 15 B4 1.769 83.02 46.15 100.00 0.632 16 B6 1.766 83.02 46.15 100.00 0.632 17 B8 1.758 83.02 46.15 100.00 0.632 18 Mean-B2 1.524 79.89 36.20 100.00 0.532
Table 6.
Comparison of monitoring accuracy of forest disturbance without feature compression and with feature compression View tableView in ArticleTable 6.
Comparison of monitoring accuracy of forest disturbance without feature compression and with feature compression 扰动类型 特征压缩 准确率/% 召回率/% 精确度/% F1 score 森林火灾 压缩前 93.35 91.36 95.48 0.934 压缩后 92.65 90.57 94.86 0.927 森林砍伐 压缩前 86.38 74.26 97.12 0.842 压缩后 87.46 75.74 98.10 0.855 森林滑坡 压缩前 95.43 74.42 72.73 0.736 压缩后 94.04 79.07 61.82 0.694
Tools
Get Citation
Copy Citation Text
Junying SONG, Xiufang ZHU, Mingxiu TANG, Rui GUO. Optimal Feature Selection for Forest Disturbance Monitoring[J]. Remote Sensing Technology and Application, 2025, 40(4): 1002
Paper Information
Category:
Received: Apr. 11, 2024
Accepted: --
Published Online: Aug. 26, 2025
The Author Email: Xiufang ZHU (zhuxiufang@bnu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20