基于Sentinel-1/2时空融合模型的冠层高度反演研究

doi:10.13474/j.cnki.11-2246.2026.0312

摘要/Abstract

摘要： 针对森林冠层高度估算中存在的实地监测效率低、成本高及单一遥感数据分辨率不足和下垫面数据缺失等问题,本文基于ConvLSTM-Transformer混合神经网络提出了一种多源时空融合深度学习模型(SST-CLT)。该模型融合 Sentinel-1 时间序列SAR数据、Sentinel-2 多光谱数据及机载 LiDAR 冠层高度真值数据,通过协同建模时间、空间和光谱特征提升反演精度。研究表明,模型训练集精度为R²=0.93,RMSE=3.10 m,MAE=2.39 m;验证集精度为R²=0.84,RMSE=4.54 m,MAE=3.00 m;总样本集精度达R²=0.89,RMSE=3.81 m,MAE=2.65 m,均表现优异;生成的冠层高度图能精准刻画空间异质性。SST-CLT模型兼具高精度与强泛化能力,可为森林资源动态监测及生态系统研究提供可靠技术支撑。

关键词: 森林冠层高度, 森林生物量, 机载LiDAR, Autokeras, SST-CLT模型

Abstract: To address the limitations in forest canopy height estimation,including low efficiency and high costs of field monitoring,insufficient resolution of single-source remote sensing data,and lack of underlying surface parameters.This study proposes a multi-source spatio-temporal fusion deep learning model named SST-CLT,based on a ConvLSTM-Transformer hybrid neural network.The model integrates Sentinel-1 time-series SAR data,Sentinel-2 multi-spectral data,and airborne LiDAR-derived canopy height reference data to enhance inversion accuracy through collaborative spatio-temporal-spectral feature modeling.The results demonstrate excellent performance: the training set achieved R²=0.93,RMSE=3.10 m,and MAE=2.39 m,while the validation set yielded R²=0.84,RMSE=4.54 m,and MAE=3.00 m.The overall sample set reached R²=0.89,RMSE=3.81 m,MAE=2.65 m,with generated canopy height maps accurately characterizing spatial heterogeneity.The SST-CLT model exhibits both high precision and strong generalization capability,providing reliable technical support for forest resource dynamic monitoring and ecosystem studies.

Key words: forest canopy height, forest biomass, airborne LiDAR, Autokeras, SST-CLT model

中图分类号:

P237

张雯雯, 周伟, 王杰. 基于Sentinel-1/2时空融合模型的冠层高度反演研究[J]. 测绘通报, 2026, 0(3): 68-74.

ZHANG Wenwen, ZHOU Wei, WANG Jie. Research on canopy height inversion based on Sentinel-1/2 spatio-temporal fusion model[J]. Bulletin of Surveying and Mapping, 2026, 0(3): 68-74.

参考文献

[1] 温丁,徐丽,王春燕,等.京津冀城区森林植被碳汇遥感评估与预测[J].地理科学进展,2024,43(12):2520-2529.
[2] 朱宁宁,杨必胜,董震,等.森林植被碳储量的遥感估测流程与方法[J].遥感学报,2025,29(1):134-146.
[3] 骆社周,王成,习晓环,等.星载激光雷达GLAS与TM光学遥感联合反演森林叶面积指数[J].红外与毫米波学报,2015,34(2):243-249.
[4] BOISVENUE C,SMILEY B P,WHITE J C,et al.Integration of Landsat time series and field plots for forest productivity estimates in decision support models[J].Forest Ecology and Management,2016,376:284-297.
[5] GHOSH A,FASSNACHT F E,JOSHI P K,et al.A framework for mapping tree species combining hyperspectral and LiDAR data:role of selected classifiers and sensor across three spatial scales[J].International Journal of Applied Earth Observation and Geoinformation,2014,26:49-63.
[6] JIN Shichao,SU Yanjun,GAO Shang,et al.The transferability of random forest in canopy height estimation from multi-source remote sensing data[J].Remote Sensing,2018,10(8):1183.
[7] KLOSTERMAN S,MELAAS E,WANG J A,et al.Fine-scale perspectives on landscape phenology from unmanned aerial vehicle (UAV)photography[J].Agricultural and Forest Meteorology,2018,248:397-407.
[8] LU Bing,HE Yuhong.Species classification using unmanned aerial vehicle (UAV)-acquired high spatial resolution imagery in a heterogeneous grassland[J].ISPRS Journal of Photogrammetry and Remote Sensing,2017,128:73-85.
[9] GHASEMI N,SAHEBI M R,MOHAMMADZADEH A.A review on biomass estimation methods using synthetic aperture radar data[J].International Journal of Geomatics and Geosciences,2023,1:776-788.
[10] HUDAK A T,LEFSKY M A,COHEN W B,et al.Integration of LiDAR and Landsat ETM+data for estimating and mapping forest canopy height[J].Remote Sensing of Environment,2002,82(2/3):397-416.
[11] 王鹏杰,张绘芳,田昕,等.主被动遥感特征优化的天然林森林蓄积量估测:以新疆巩留县为例[J].遥感技术与应用,2022,37(3):672-680.
[12] LI Wang,NIU Zheng,SHANG Rong,et al.High-resolution mapping of forest canopy height using machine learning by coupling ICESat-2 LiDAR with Sentinel-1,Sentinel-2 and Landsat 8 data[J].International Journal of Applied Earth Observation and Geoinformation,2020,92:102163.
[13] LANG N,SCHINDLER K,WEGNER J D.Country-wide high-resolution vegetation height mapping with Sentinel-2[J].Remote Sensing of Environment,2019,233:111347.
[14] SHI X,CHEN Z,WANG H,et al.Convolutional LSTM network:a machine learning approach for precipitation nowcasting[J].Advances in Neural Information Processing Systems,2015,28(15):135-165.
[15] ALI SHAH S A,MANZOOR M A,BAIS A.Canopy height estimation at landsat resolution using convolutional neural networks[J].Machine Learning and Knowledge Extraction,2020,2(1):23-36.
[16] TORRES DE ALMEIDA C,GERENTE J,RODRIGO DOS PRAZERES CAMPOS J,et al.Canopy height mapping by Sentinel-1 and 2 satellite images,airborne LiDAR data,and machine learning[J].Remote Sensing,2022,14(16):4112.
[17] HANSEN M C,POTAPOV P V,GOETZ S J,et al.Mapping tree height distributions in Sub-Saharan Africa using Landsat 7 and 8 data[J].Remote Sensing of Environment,2016,185:221-232.
[18] WANG Yuanyuan,LI Guicai,DING Jianhua,et al.A combined GLAS and MODIS estimation of the global distribution of mean forest canopy height[J].Remote Sensing of Environment,2016,174:24-43.
[19] PHAM L T H,BRABYN L.Monitoring mangrove biomass change in Vietnam using SPOT images and an object-based approach combined with machine learning algorithms[J].ISPRS Journal of Photogrammetry and Remote Sensing,2017,128:86-97.
[20] HE Xin,ZHOU Yong,ZHAO Jiaqi,et al.Swin transformer embedding UNet for remote sensing image semantic segmentation[J].IEEE Transactions on Geoscience and Remote Sensing,2022,60(5):1-15.
[21] EIGEN D,PUHRSCH C,FERGUS R.Depth map prediction from a single image using a multi-scale deep network[J].Advances In Neural Information Processing Systems,2014,2(9):15-19.