基于Sentinel-1和Sentinel-2的多特征组合下的火烧迹地变化检测

doi:10.12118/j.issn.1000-6060.2025.244

干旱区地理 ›› 2026, Vol. 49 ›› Issue (4): 791-803.doi: 10.12118/j.issn.1000-6060.2025.244 cstr: 32274.14.ALG2025244

基于Sentinel-1和Sentinel-2的多特征组合下的火烧迹地变化检测

安佳乐¹^,²(), 王蕾¹^,³, 郑江华¹^,²(), 李建豪¹^,², 赵国兵¹^,², 尼格拉·吐尔逊¹^,², 伏蓉¹^,², 罗磊³

¹ 新疆大学地理与遥感科学学院，新疆乌鲁木齐 830046
² 新疆大学绿洲重点实验室，新疆乌鲁木齐 830046
³ 新疆林业科学院现代林业研究所，新疆乌鲁木齐 830000

收稿日期:2025-05-06 修回日期:2025-05-26 出版日期:2026-04-25 发布日期:2026-04-28
通讯作者: 郑江华（1973-），男，博士，教授，主要从事干旱区灾害监测、预警与损失评估建模等方面的研究. E-mail: zheng.jianghua@xju.edu.cn
作者简介:安佳乐（2000-），男，硕士研究生，主要从事深度学习与野火监测等方面的研究. E-mail: anjiale@stu.xju.edu.cn
基金资助:
新疆维吾尔自治区天山英才培养计划(2023SNGGGGCC004);新疆天山雪松计划项目(2020XS04)

Sentinel-1 and Sentinel-2 based change detection in fire trails under multi-feature combination

AN Jiale¹^,²(), WANG Lei¹^,³, ZHENG Jianghua¹^,²(), LI Jianhao¹^,², ZHAO Guobing¹^,², Nigela TUERXUN¹^,², FU Rong¹^,², LUO Lei³

¹ College of Geography and Remote Sensing Sciences, Xinjiang University, Urumqi 830046, Xinjiang, China
² Xinjiang Key Laboratory of Oasis Ecology, Xinjiang University, Urumqi 830046, Xinjiang, China
³ Institute of Modern Forestry, Xinjiang Academy of Forestry Sciences, Urumqi 830000, Xinjiang, China

Received:2025-05-06 Revised:2025-05-26 Published:2026-04-25 Online:2026-04-28

摘要/Abstract

摘要：

在全球气候变暖加剧的背景下，极端高温事件频发，导致干旱加剧和可燃物积累，显著提升了野火发生的风险。日益频繁的野火对生态安全与可持续发展构成严重威胁。目前，火烧迹地识别方法主要依赖单一光谱特征，易受植被物候、地形及大气条件干扰，存在误判和漏判问题。为此，提出通过多特征协同识别策略，旨在提升火烧迹地检测的精度，为精准防控野火提供技术支持。基于随机森林和XGBoost模型计算的SHAP特征重要性，融合多种光谱指数特征、极化特征和原始特征，设计了20种特征组合。结合火烧前后双时相Sentinel-2多光谱与Sentinel-1极化遥感影像，使用基于Transformer的双时态图像变化检测模型（BIT_CD）对中国及周边地区的火烧迹地进行识别。结果表明：（1）改进后的模型在测试集上整体准确率（OA）达到92.7%，较原始模型提升2.36%，交并比（IoU）提升5%~14%，显著增强了模型对火烧区域的识别能力。（2）综合模型的各个指标，确定了最优特征组合为改进型燃烧指数（NBR_PLUS）、Sentinel-1后向散射系数（VV、VH）、红边波段（RE4）及Sentinel-2短波红外波段（S2、S1）的融合特征。研究提出的多特征融入BIT_CD模型能够有效识别火烧迹地，为火灾灾后灾情识别、过火面积评估、灾后重建、生态恢复和可持续发展提供方法参考。

关键词: 火烧迹地, 变化检测, 特征组合, 深度学习, Transformer

Abstract:

With the increasing intensification of global warming, extremely high-temperature events occur more frequently, resulting in a considerable rise in drought conditions and the accumulation of combustible materials, thereby substantially increasing wildfire risk. The intensity and frequency of wildfires continue to increase, posing serious threats to human safety, ecological balance, and economic development. Previous studies on burn scar change detection primarily rely on independent spectral index thresholds or original spectral features. However, these approaches are highly susceptible to interference from complex environmental factors, including vegetation phenological variations, topographic shadows, and atmospheric conditions, which often lead to false detections and missed burn scar identification. Moreover, existing studies show limitations in leveraging multifeature combinations for burn scar detection, as most methods depend on single-feature extraction and lack systematic evaluation of feature synergy. To address these limitations, this study explores multifeature combination recognition methods to improve the accuracy and reliability of burn scar identification. Based on feature importance derived from SHAP using Random Forest and XGBoost models, multispectral indices, polarization, and original spectral features are integrated to construct 20 feature combinations. Using bitemporal Sentinel-2 multispectral and Sentinel-1 polarization remote sensing images acquired before and after wildfire events, the bitemporal image transformer for change detection (BIT_CD), a transformer-based bitemporal change detection model, is applied to identify burn scars in China and surrounding regions. Experimental results show that the improved model achieves an overall accuracy of 92.7% on the test set, representing a 2.36% improvement over the original model. In comparison, the Intersection over Union increased by 5%-14%, substantially enhancing burned-area detection performance. Based on comprehensive evaluation metrics, the optimal feature combination consists of the improved burn index (NBR_PLUS), Sentinel-1 backscattering coefficients (VV and VH), the red-edge band (RE4), and Sentinel-2 shortwave infrared bands (S2 and S1). The proposed multifeature integration BIT_CD model effectively identifies burn scars and provides a robust methodological reference for post-fire disaster assessment, burned-area mapping, post-disaster reconstruction, ecological restoration, and sustainable development.

Key words: fire trails, change detection, feature combination, deep learning, Transformer

安佳乐, 王蕾, 郑江华, 李建豪, 赵国兵, 尼格拉·吐尔逊, 伏蓉, 罗磊. 基于Sentinel-1和Sentinel-2的多特征组合下的火烧迹地变化检测[J]. 干旱区地理, 2026, 49(4): 791-803.

AN Jiale, WANG Lei, ZHENG Jianghua, LI Jianhao, ZHAO Guobing, Nigela TUERXUN, FU Rong, LUO Lei. Sentinel-1 and Sentinel-2 based change detection in fire trails under multi-feature combination[J]. Arid Land Geography, 2026, 49(4): 791-803.

图/表 13

表1

表2

图1

表3

图2

图3

图4

表4

表5

图5

图6

图7

表6

参考文献 55

[1]	Hamilton D, Brothers K, McCall C, et al. Mapping forest burn extent from hyperspatial imagery using machine learning[J]. Remote Sensing, 2021, 13(19): 3843, doi: 10.3390/rs13193843.
[2]	Cunningham C X, Williamson G J, Bowman D M J S. Increasing frequency and intensity of the most extreme wildfires on earth[J]. Nature Ecology & Evolution, 2024, 8: 1420-1425.
[3]	Rahman R A, White B, Ma C B. The effect of growth, deforestation, forest fires, and volcanoes on Indonesian regional air quality[J]. Journal of Cleaner Production, 2024, 457: 142311, doi: 10.1016/j.jclepro.2024.142311.
[4]	Yang W L, Wu M S, Kong L, et al. A spatial weight sampling method integrating the spatiotemporal pattern enhances the understanding of the occurrence mechanism of wildfires in the southwestern mountains of China[J]. Forest Ecology and Management, 2025, 585: 122619, doi: 10.1016/j.foreco.2025.122619.
[5]	Qin Y J, Wang H C, Wang Y M, et al. Wildfires in southeast Asia pollute the atmosphere in the northern south China sea[J]. Science Bulletin, 2024, 69(8): 1011-1015. doi: 10.1016/j.scib.2024.02.026 pmid: 38431468
[6]	夏文浩, 霍瑜, 逯渊, 等. 新疆农业碳排放的时空差异与空间溢出效应分析[J]. 干旱区地理, 2024, 47(6): 1084-1096. doi: 10.12118/j.issn.1000-6060.2023.344
	[Xia Wenhao, Huo Yu, Lu Yuan, et al. Spatialtemporal differences and spatial spillover effects of agricultural carbon emissions in Xinjiang[J]. Arid Land Geography, 2024, 47(6): 1084-1096.] doi: 10.12118/j.issn.1000-6060.2023.344
[7]	Hethcoat M G, Carreiras J M B, Edwards D P, et al. Detecting tropical selective logging with C-band SAR data may require a time series approach[J]. Remote Sensing of Environment, 2021, 259: 112411, doi: 10.1016/j.rse.2021.112411.
[8]	Hethcoat M G, Edwards D P, Carreiras J M B, et al. A machine learning approach to map tropical selective logging[J]. Remote Sensing of Environment, 2019, 221: 569-582. doi: 10.1016/j.rse.2018.11.044
[9]	Chen W T, Li X J, He H X, et al. A review of fine-scale land use and land cover classification in open-pit mining areas by remote sensing techniques[J]. Remote Sensing, 2018, 10(1): 15, doi: 10.3390/rs10010015.
[10]	Chen Y, Morton D C, Randerson J T. Remote sensing for wildfire monitoring: Insights into burned area, emissions, and fire dynamics[J]. One Earth, 2024, 7(6): 1022-1028. doi: 10.1016/j.oneear.2024.05.014
[11]	Schroeder W, Oliva P, Giglio L, et al. Active fire detection using Landsat-8/OLI data[J]. Remote Sensing of Environment, 2016, 185: 210-220. doi: 10.1016/j.rse.2015.08.032
[12]	Radeloff V C, Roy D P, Wulder M A, et al. Need and vision for global medium-resolution Landsat and Sentinel-2 data products[J]. Remote Sensing of Environment, 2024, 300: 113918, doi: 10.1016/j.rse.2023.113918.
[13]	Bastarrika A, Rodriguez-Montellano A, Roteta E, et al. An automatic procedure for mapping burned areas globally using Sentinel-2 and VIIRS/MODIS active fires in Google Earth Engine[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2024, 218: 232-245. doi: 10.1016/j.isprsjprs.2024.08.019
[14]	Lin Z Y, Chen F, Niu Z, et al. An active fire detection algorithm based on multi-temporal FengYun-3C VIRR data[J]. Remote Sensing of Environment, 2018, 211: 376-387. doi: 10.1016/j.rse.2018.04.027
[15]	Kang Y, Sung T, Im J. Toward an adaptable deep-learning model for satellite-based wildfire monitoring with consideration of environmental conditions[J]. Remote Sensing of Environment, 2023, 298: 113814, doi: 10.1016/j.rse.2023.113814.
[16]	曾洪涛. 基于 GK2A 卫星数据的野火实时监测方法[D]. 成都: 电子科技大学, 2024.
	[Zeng Hongtao. Near real-time wildfire monitoring methods based on GK2A data[D]. Chengdu: University of Electronic Science and Technology of China, 2024.]
[17]	饶月明, 王川, 黄华国. 联合多源遥感数据监测四川木里县森林火灾[J]. 遥感学报, 2020, 24(5): 559-570.
	[Rao Yueming, Wang Chuan, Huang Huaguo. Forest fire monitoring based on multisensor remote sensing techniques in Muli County, Sichuan Province[J]. Journal of Remote Sensing, 2020, 24(5): 559-570.]
[18]	Meneses B M. Vegetation recovery patterns in burned areas assessed with Landsat 8 OLI imagery and environmental biophysical data[J]. Fire, 2021, 4(4): 76, doi: 10.3390/fire4040076.
[19]	López G M J, Caselles V. Mapping burns and natural reforestation using thematic mapper data[J]. Geocarto International, 1991, 6(1): 31-37.
[20]	Rouse J W, Haas R H, Schell J A, et al. Monitoring the vernal advancement and retrogradation (green wave effect) of natural vegetation[R]. Greenbelt: NASA/GSFC, 1974.
[21]	Pettorelli N, Vik J O, Mysterud A, et al. Using the satellite-derived NDVI to assess ecological responses to environmental change[J]. Trends in Ecology & Evolution, 2005, 20(9): 503-510. doi: 10.1016/j.tree.2005.05.011
[22]	Xu H Q, Chen J H, He G J, et al. Immediate assessment of forest fire using a novel vegetation index and machine learning based on multi-platform, high temporal resolution remote sensing images[J]. International Journal of Applied Earth Observation and Geoinformation, 2024, 134: 104210, doi: 10.1016/j.jag.2024.104210.
[23]	卢涛, 张军, 胡文英. 基于Sentinel-2A/B数据的火烧迹地提取方法研究[J]. 三峡生态环境监测, 2021, 6(2): 52-60.
	[Lu Tao, Zhang Jun, Hu Wenying. Extraction method of burned area based on Sentinel-2A/B data[J]. Ecology and Environmental Monitoring of Three Gorges, 2021, 6(2): 52-60.]
[24]	Zhang P Z, Ban Y F, Nascetti A. Total-variation regularized U-Net for wildfire burned area mapping based on Sentinel-1 C-band SAR backscattering data[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2023, 203: 301-313. doi: 10.1016/j.isprsjprs.2023.07.024
[25]	Ban Y F, Zhang P Z, Nascetti A, et al. Near real-time wildfire progression monitoring with Sentinel-1 SAR time series and deep learning[J]. Scientific Reports, 2020, 10(1): 1322, doi: 10.1038/s41598-019-56967-x. pmid: 31992723
[26]	Stavrakoudis D, Katagis T, Minakou C, et al. Automated burned scar mapping using sentinel-2 imagery[J]. Journal of Geographic Information System, 2020, 12(3): 221-240. doi: 10.4236/jgis.2020.123014
[27]	Tariq A, Jiango Y, Lu L L, et al. Integrated use of sentinel-1 and sentinel-2 data and open-source machine learning algorithms for burnt and unburnt scars[J]. Geomatics, Natural Hazards and Risk, 2023, 14(1): 2190856, doi: 10.1080/19475705.2023.2190856.
[28]	Gorelick N, Hancher M, Dixon M, et al. Google earth engine: Planetary-scale geospatial analysis for everyone[J]. Remote Sensing of Environment, 2017, 202: 18-27. doi: 10.1016/j.rse.2017.06.031
[29]	LeCun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521(7553): 436-444. doi: 10.1038/nature14539
[30]	Liu P, Liu Y X, Guo X X, et al. Burned area detection and mapping using time series Sentinel-2 multispectral images[J]. Remote Sensing of Environment, 2023, 296: 113753, doi: 10.1016/j.rse.2023.113753.
[31]	Sui T, Huang Q Y, Wu M D, et al. BiAU-Net: Wildfire burnt area mapping using bi-temporal Sentinel-2 imagery and U-Net with attention mechanism[J]. International Journal of Applied Earth Observation and Geoinformation, 2024, 132: 104034, doi: 10.1016/j.jag.2024.104034.
[32]	Zhao F, Sun R, Zhong L H, et al. Monthly mapping of forest harvesting using dense time series Sentinel-1 SAR imagery and deep learning[J]. Remote Sensing of Environment, 2022, 269: 112822, doi: 10.1016/j.rse.2021.112822.
[33]	Chen H, Qi Z, Shi Z. Remote sensing image change detection with Transformers[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1-14.
[34]	Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]// Neural Information Processing Systems. Proceedings of the 31st International Conference on Neural Information Processing Systems. New York: Curran Associates Inc., 2017: 6000-6010.
[35]	Ronneberger O, Fischer P, Brox T. U-Net: Convolutional networks for biomedical image segmentation[J]. arXiv, 2015, doi: 10.48550/arXiv.1505.04597.
[36]	Collins L, Griffioen P, Newell G, et al. The utility of random forests for wildfire severity mapping[J]. Remote Sensing of Environment, 2018, 216: 374-384. doi: 10.1016/j.rse.2018.07.005
[37]	方创琳, 刘海猛, 罗奎, 等. 中国人文地理综合区划[J]. 地理学报, 2017, 72(2): 179-196. doi: 10.11821/dlxb201702001
	[Fang Chuanglin, Liu Haimeng, Luo Kui, et al. Comprehensive regionalization of human geography in China[J]. Acta Geographica Sinica, 2017, 72(2): 179-196.] doi: 10.11821/dlxb201702001
[38]	Feurdean A, Florescu G, Tanţău I, et al. Recent fire regime in the southern boreal forests of western siberia is unprecedented in the last five millennia[J]. Quaternary Science Reviews, 2020, 244: 106495, doi: 10.1016/j.quascirev.2020.106495.
[39]	金相皓, 李鹏, 李霞, 等. 基于CHELSA的1980—2018年南亚——东南亚气候变化特征与趋势分析[J]. 地理科学进展, 2023, 42(7): 1365-1379. doi: 10.18306/dlkxjz.2023.07.011
	[Jin Xianghao, Li Peng, Li Xia, et al. A CHELSA-based analysis of climate change characteristics and trends in south and southeast Asia (SSEA) during 1980—2018[J]. Progress in Geography, 2023, 42(7): 1365-1379.] doi: 10.18306/dlkxjz.2023.07.011
[40]	杨晓颖, 玉山, 都瓦拉, 等. 蒙古高原草原火灾风险评价研究[J]. 干旱区地理, 2021, 44(4): 1032-1044. doi: 10.12118/j.issn.1000–6060.2021.04.16
	[Yang Xiaoying, Yu Shan, Duwala, et al. Risk assessment of grassland fire on the Mongolian Plateau[J]. Arid Land Geography, 2021, 44(4): 1032-1044.] doi: 10.12118/j.issn.1000–6060.2021.04.16
[41]	Zheng B, Ciais P, Chevallier F, et al. Record-high CO₂ emissions from boreal fires in 2021[J]. Science, 2023, 379(6635): 912-917. doi: 10.1126/science.ade0805 pmid: 36862792
[42]	何阳, 杨进, 马勇, 等. 基于Landsat-8陆地卫星数据的火点检测方法[J]. 红外与毫米波学报, 2016, 35(5): 600-609.
	[He Yang, Yang Jin, Ma Yong, et al. A method for fire detection using Landsat 8 data[J]. Journal of Infrared and Millimeter Waves, 2016, 35(5): 600-609.]
[43]	Roy D P, Huang H Y, Boschetti L, et al. Landsat-8 and Sentinel-2 burned area mapping-A combined sensor multi-temporal change detection approach[J]. Remote Sensing of Environment, 2019, 231: 111254, doi: 10.1016/j.rse.2019.111254.
[44]	De Almeida Pereira G H, Fusioka A M, Nassu B T, et al. Active fire detection in Landsat-8 imagery: A large-scale dataset and a deep-learning study[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2021, 178: 171-186. doi: 10.1016/j.isprsjprs.2021.06.002
[45]	Pulvirenti L, Squicciarino G, Fiori E, et al. Near real-time generation of a country-level burned area database for Italy from Sentinel-2 data and active fire detections[J]. Remote Sensing Applications: Society and Environment, 2023, 29: 100925, doi: 10.1016/j.rsase.2023.100925.
[46]	周乐民. 基于多源遥感影像的森林火场信息提取与应用——以“4.20冕宁县森林火灾”为例[D]. 成都: 西南交通大学, 2022.
	[Zhou Lemin. Extraction and application of forest fire information based on multi source remote sensing imagery: A case study of “4.20 mianning county forest fire”[D]. Chengdu: Southwest Jiaotong University, 2022.]
[47]	Pasquarella V J, Brown C F, Czerwinski W, et al. Comprehensive quality assessment of optical satellite imagery using weakly supervised video learning[C]// IEEE/CVF. Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). New York: IEEE. 2023: 2125-2135.
[48]	Russell B C, Torralba A, Murphy K P, et al. LabelMe: A database and web-based tool for image annotation[J]. International Journal of Computer Vision, 2008, 77(1): 157-173. doi: 10.1007/s11263-007-0090-8
[49]	Breiman L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32. doi: 10.1023/A:1010933404324
[50]	Chen T Q, Guestrin C. XGBoost: A scalable tree boosting system[C]// Proceedings of the 22nd ACM Sigkdd International Conference on Knowledge Discovery and Data Mining. New York: Association for Computing Machinery, 2016: 785-794.
[51]	赵国兵, 郑江华, 王蕾, 等. 基于RF分类调优和SNIC聚类的新疆红枣种植区遥感提取[J]. 干旱区地理, 2024, 47(6): 1004-1014. doi: 10.12118/j.issn.1000-6060.2023.382
	[Zhao Guobing, Zheng Jianghua, Wang Lei, et al. Remote sensing extraction of jujube planting area in Xinjiang based on RF classification optimization and SNIC clustering[J]. Arid Land Geography, 2024, 47(6): 1004-1014.] doi: 10.12118/j.issn.1000-6060.2023.382
[52]	Genuer R, Poggi J M, Tuleau-Malot C. Variable selection using random forests[J]. Pattern Recognition Letters, 2010, 31(14): 2225-2236. doi: 10.1016/j.patrec.2010.03.014
[53]	Verrelst J, Muñoz J, Alonso L, et al. Machine learning regression algorithms for biophysical parameter retrieval: Opportunities for sentinel-2 and -3[J]. Remote Sensing of Environment, 2012, 118: 127-139. doi: 10.1016/j.rse.2011.11.002
[54]	Friedman J H. Greedy function approximation: A gradient boosting machine[J]. Annals of Statistics, 2001, 29(5): 1189-1232. doi: 10.1214/aos/1013203450
[55]	Bergstra J, Bardenet R, Bengio Y, et al. Algorithms for hyper-parameter optimization[C]// Neural Information Processing Systems. Proceedings of the 25th International Conference on Neural Information Processing Systems. New York: Curran Associates Inc., 2011: 2546-2554.

光谱波段	空间分辨率/m	名称	简称
Blue	10	蓝波段	B
Green	10	绿波段	G
Red	10	红波段	R
RE1	10	红边1波段	RE1
RE2	10	红边2波段	RE2
RE3	10	红边3波段	RE3
NIR	10	近红外波段	NIR
RE4	10	红边4波段	RE4
SWIR1	20	短波红外1波段	S1
SWIR2	20	短波红外2波段	S2

光谱波段	空间分辨率/m	名称	极化方式
VV	10	垂直发射-垂直接收	同极化
VH	10	垂直发射-水平接收	交叉极化

简称	光谱指数
EVI	增强型植被指数
NDVI	归一化植被指数
IBI	生物完整性指数
BSI	裸土指数
SSI	土壤表面指数
RVI	比率植被指数
DVI	差异植被指数
GCVI	绿色叶绿素植被指数
NBR	归一化燃烧指数
NBR2	改进的归一化燃烧指数
BAI	燃烧区域指数
MIRBI	中红外燃烧指数
NBR_PLUS	增强型归一化燃烧指数
BAIS2	燃烧区域指数
REP	红边位置指数

编号	植被指数	Sentinel-1	原始波段	组合方式
1	-	-	RE4、S2、RE2、S1、RE1、RE3	GEE_RF
2	-	-	RE4、RE1、S1、RE3、NIR、S2	RF
3	-	-	RE4、S2、S1、RE1、G、R	XGBoost
4	-	VV、VH	RE4、S2、S1、RE1	无光谱指数
5	MIRBI、NBR_PLUS	-	RE4、S2、S1、RE1	无Sentinel-1
6	MIRBI、NBR2	-	RE4、S2、S1、RE1	无Sentinel-1
7	NBR2、NBR_PLUS	-	RE4、S2、S1、RE1	无Sentinel-1
8	MIRBI	VV、VH	RE4、S1、RE1	一种光谱指数
9	MIRBI	VV、VH	RE4、S2、RE1	一种光谱指数
10	MIRBI	VV、VH	RE4、S2、S1	一种光谱指数
11	MIRBI	VV、VH	S2、S1、RE1	一种光谱指数
12	NBR_PLUS	VV、VH	RE4、S1、RE1	一种光谱指数
13	NBR_PLUS	VV、VH	RE4、S2、RE1	一种光谱指数
14	NBR_PLUS	VV、VH	RE4、S2、S1	一种光谱指数
15	NBR_PLUS	VV、VH	S2、S1、RE1	一种光谱指数
16	NBR2	VV、VH	RE4、S1、RE1	一种光谱指数
17	NBR2	VV、VH	RE4、S2、RE1	一种光谱指数
18	NBR2	VV、VH	RE4、S2、S1	一种光谱指数
19	NBR2	VV、VH	S2、S1、RE1	一种光谱指数
20	-	-	R、G、B	RGB

特征组合方式	OA	IoU	F1	Precision	Recall
GEE_RF	92.09	53.82	69.98	73.57	66.72
RF	92.37	56.29	72.04	73.00	71.10
XGBoost	92.71	56.52	72.22	76.24	68.61
VV_VH_RE4_S2_S1_RE1	91.67	49.62	66.33	75.14	59.37
MIRBI_NBR_PLUS_RE4_S2_S1_RE1	92.67	56.09	71.87	76.56	67.72
MIRBI_NBR2_RE4_S2_S1_RE1	92.13	51.67	68.13	77.31	60.90
NBR2_NBR_PLUS_RE4_S2_S1_RE1	92.31	52.85	69.15	77.60	62.36
MIRBI_VV_VH_RE4_S1_RE1	92.12	56.03	71.82	71.03	72.63
MIRBI_VV_VH_RE4_S2_RE1	91.86	52.86	69.16	72.58	66.05
MIRBI_VV_VH_RE4_S2_S1	92.76	58.00	73.42	74.53	72.34
MIRBI_VV_VH_S2_S1_RE1	92.07	53.03	69.31	74.53	64.77
NBR_PLUS_VV_VH_RE4_S1_RE1	92.29	53.64	69.83	76.07	64.53
NBR_PLUS_VV_VH_RE4_S2_RE1	92.09	55.29	71.21	71.66	70.75
NBR_PLUS_VV_VH_RE4_S2_S1	92.92	57.62	73.11	76.95	69.64
NBR_PLUS_VV_VH_S2_S1_RE1	92.50	56.10	71.87	74.64	69.31
NBR2_VV_VH_RE4_S1_RE1	91.35	50.24	66.88	71.03	63.18
NBR2_VV_VH_RE4_S2_RE1	91.49	49.46	66.18	73.43	60.24
NBR2_VV_VH_RE4_S2_S1	91.96	51.52	68.00	75.51	61.86
NBR2_VV_VH_S2_S1_RE1	92.17	50.15	66.80	80.64	57.02
RGB	90.56	44.24	61.34	68.30	55.68

基于Sentinel-1和Sentinel-2的多特征组合下的火烧迹地变化检测

Sentinel-1 and Sentinel-2 based change detection in fire trails under multi-feature combination

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 55

相关文章 5

Metrics

本文评价

推荐阅读 0

特征组合方式	OA	IoU	F1	Precision	Recall
GEE_RF	93.79	52.24	58.54	64.16	61.20
RF	94.71	57.94	69.35	83.04	71.29
XGBoost	94.39	63.58	79.44	77.22	77.41
VV_VH_RE4_S2_S1_RE1	90.28	43.74	50.07	64.25	53.67
MIRBI_NBR_PLUS_RE4_S2_S1_RE1	93.88	57.51	76.65	75.57	70.37
MIRBI_NBR2_RE4_S2_S1_RE1	92.70	56.95	61.32	90.77	70.64
NBR2_NBR_PLUS_RE4_S2_S1_RE1	95.53	55.50	84.15	67.69	65.93
MIRBI_VV_VH_RE4_S1_RE1	90.77	38.37	43.52	66.92	46.12
MIRBI_VV_VH_RE4_S2_RE1	91.28	41.13	68.87	70.21	48.47
MIRBI_VV_VH_RE4_S2_S1	82.41	26.89	46.84	48.40	36.67
MIRBI_VV_VH_S2_S1_RE1	92.16	41.92	44.06	68.93	50.43
NBR_PLUS_VV_VH_RE4_S1_RE1	86.13	40.12	59.91	53.37	51.95
NBR_PLUS_VV_VH_RE4_S2_RE1	82.81	31.45	53.05	45.32	42.30
NBR_PLUS_VV_VH_RE4_S2_S1	95.92	71.87	81.60	85.55	83.39
NBR_PLUS_VV_VH_S2_S1_RE1	92.90	54.07	60.59	65.87	62.76
NBR2_VV_VH_RE4_S1_RE1	82.12	16.85	42.88	19.18	23.02
NBR2_VV_VH_RE4_S2_RE1	88.25	33.10	44.24	58.26	41.98
NBR2_VV_VH_RE4_S2_S1	91.38	38.91	44.82	64.28	48.00
NBR2_VV_VH_S2_S1_RE1	93.00	42.64	65.87	59.63	52.74
RGB	86.63	33.23	77.59	46.32	46.59

[1]	周灵敏, 伊木然·库鲁万, 玉素甫江·如素力, 吴海智, 娜扎开提·尼加提. 中国西北干旱区大型湖泊湖冰物候时空差异及影响因素[J]. 干旱区地理, 2026, 49(3): 484-495.
[2]	王圣节, 张庆红, 桑铭键. 基于改进深度学习模型的干旱区城市PM_2.5浓度预测与应用研究[J]. 干旱区地理, 2026, 49(2): 343-355.
[3]	郭佳丽, 马勇刚, 潘恒, 李娜, 孙长宁, 孙倩, 周文昶, 党瑀璇. 艾比湖春夏季土壤盐渍化卫星监测对比分析[J]. 干旱区地理, 2025, 48(12): 2143-2157.
[4]	袁盼丽,汪传建,赵庆展,王学文,任媛媛,杨启原. 基于深度学习的寒旱区多时序影像土地利用及变化监测——以新疆莫索湾垦区为例[J]. 干旱区地理, 2021, 44(6): 1717-1728.
[5]	卢文路, 刘志锋, 何春阳, 夏沛. 基于Sentinel-1A合成孔径雷达数据和全卷积网络的城市建设用地监测方法研究 [J]. 干旱区地理, 2020, 43(3): 750-760.