Précédent|Prochain

Classification forestière basée sur la fusion par attention croisée des données multispectrales et SAR

  • role： First author第一作者
  • Affiliation：

    State Key Laboratory of Efficient Production of Forest Resources, Beijing Forestry University, Beijing 100083, China

    Beijing Key Laboratory of Precision Forestry, College of Forestry, Beijing Forestry University, Beijing 100083, China

    Key Laboratory of Forest Cultivation and Protection, Ministry of Education, Beijing Forestry University, Beijing 100083, China

  • Email：18801495791@163.com
  • Introduction：谢一帆，研究方向为林业遥感与信息技术。E-mail： 18801495791@163.com
XIE Yifan,  
  • Affiliation：

    State Key Laboratory of Efficient Production of Forest Resources, Beijing Forestry University, Beijing 100083, China

    Beijing Key Laboratory of Precision Forestry, College of Forestry, Beijing Forestry University, Beijing 100083, China

    Key Laboratory of Forest Cultivation and Protection, Ministry of Education, Beijing Forestry University, Beijing 100083, China

JIA Zihan,  
  • role： Corresponding author通信作者
  • Affiliation：

    State Key Laboratory of Efficient Production of Forest Resources, Beijing Forestry University, Beijing 100083, China

    Beijing Key Laboratory of Precision Forestry, College of Forestry, Beijing Forestry University, Beijing 100083, China

    Key Laboratory of Forest Cultivation and Protection, Ministry of Education, Beijing Forestry University, Beijing 100083, China

  • Email：zhangxl@bjfu.edu.cn
  • Introduction：张晓丽，研究方向为林业遥感与信息技术。E-mail： zhangxl@bjfu.edu.cn
ZHANG Xiaoli*

résumé

La technologie de télédétection, grâce à sa large couverture, sa haute réactivité temporelle et sa capacité à obtenir des informations multidimensionnelles, est devenue un outil important pour les enquêtes forestières. Les images multispectrales possèdent une résolution spatiale et spectrale élevée, permettant de capturer efficacement les différences spectrales entre différents types de couvert végétal, tandis que les données radar à synthèse d'ouverture (SAR) fournissent des informations stables sur la structure de surface et les caractéristiques de texture, constituant un complément important aux caractéristiques spectrales. Cependant, les différences de structure modale et d'expression d'information entre les données actives et passives entraînent souvent des effets de fusion limités, affectant la précision de classification. Pour résoudre ce problème, cette étude a pris comme zone d'étude une partie de la ville de Pu'er dans la province du Yunnan, sélectionnant 3 types d'espèces forestières (groupes) (Pinus kesiya, eucalyptus, chêne) et un type de forêt économique (plantation de thé) ainsi que 3 autres types de couverture terrestre pour la classification, en construisant un réseau d'apprentissage profond de classification forestière fusionnant les données de télédétection actives et passives sous condition d'échantillons réduits. La méthode fusionne les images multispectrales Sentinel-2 et les images SAR Sentinel-1, propose un réseau de fusion d'attention croisée principalement basé sur les images multispectrales, et introduit un mécanisme de régularisation à portes avec taux de dropout apprenants pour permettre une fusion dynamique des caractéristiques d'auto-attention et d'attention croisée. Lors du processus de fusion, le modèle extrait les caractéristiques saillantes des images multispectrales via le mécanisme d'auto-attention, combiné à l'attention croisée guidant la réponse aux zones clés SAR, contrôlant dynamiquement la contribution des différentes modalités lors de la fusion des caractéristiques. Les résultats expérimentaux montrent que la méthode proposée obtient les meilleures performances parmi plusieurs stratégies et niveaux de fusion, avec une précision globale de classification de 95,24%, et des précisions de classification pour l'eucalyptus, les plantations de thé, le chêne et le pin kesiya respectivement de 96,78%, 94,07%, 91,73% et 92,90%. Cette étude confirme l'efficacité du mécanisme d'attention croisée dans la modélisation collaborative des informations actives et passives de télédétection, fournissant une approche réalisable et un support technique pour la classification forestière multi-source en environnements complexes.

mots-clés

Sentinel-2;Sentinel-1;attention croisée;fusion de caractéristiques;classification forestière;réseau prototype;Concrete Dropout

References

  1. 1.
    Adrian J, Sagan V and Maimaitijiang M. 2021. Sentinel SAR-optical fusion for crop type mapping using deep learning and Google Earth Engine. ISPRS Journal of Photogrammetry and Remote Sensing, 175: 215-235
  2. 2.
    Arasumani M, Bunyan M and Robin V V. 2021. Opportunities and challenges in using remote sensing for invasive tree species management, and in the identification of restoration sites in tropical montane grasslands. Journal of Environmental Management, 280: 111759
  3. 3.
    Bahdanau D, Cho K and Bengio Y. 2016. Neural machine translation by jointly learning to align and translate. arXiv preprint arXiv:1409.0473
  4. 4.
    Bjerreskov K S, Nord-Larsen T and Fensholt R. 2021. Classification of nemoral forests with fusion of multi-temporal sentinel-1 and 2 data. Remote Sensing, 13(5): 950
  5. 5.
    Boudiaf M, Masud Z I, Rony J, Dolz J, Piantanida P and Ben Ayed I. 2020. Information maximization for few-shot learning. arXiv:2008.11297
  6. 6.
    Chen B, Huang B and Xu B. 2017. Multi-source remotely sensed data fusion for improving land cover classification. ISPRS Journal of Photogrammetry and Remote Sensing, 124: 27-39
  7. 7.
    Chen C, He X Y, Guo B Y, Zhao X and Chu Y L. 2020. A pixel-level fusion method for multi-source optical remote sensing image combining the principal component analysis and curvelet transform. Earth Science Informatics, 13(4): 1005-1013
  8. 8.
    Chen L, Tian X M, Chai G Q, Zhang X L and Chen E. 2021. A new CBAM-P-net model for few-shot forest species classification using airborne hyperspectral images. Remote Sensing, 13(7): 1269
  9. 9.
    Dai Y M, Gieseke F, Oehmcke S, Wu Y Q and Barnard K. 2021. Attentional feature fusion//Proceedings of the 2021 IEEE Winter Conference on Applications of Computer Vision. Waikoloa: IEEE: 3559-3568
  10. 10.
    Dubois C, Mueller M M, Pathe C, Jagdhuber T, Cremer F, Thiel C and Schmullius C. 2020. Characterization of land cover seasonality in sentinel-1 time series data. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, V-3-2020: 97-104
  11. 11.
    Gal Y, Hron J and Kendall A. 2017. Concrete dropout//Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach: Curran Associates Inc.: 3584-3593
  12. 12.
    Gao J, Li P, Chen Z K and Zhang J N. 2020. A survey on deep learning for multimodal data fusion. Neural Computation, 32(5): 829-864
  13. 13.
    Ghassemian H. 2016. A review of remote sensing image fusion methods. Information Fusion, 32: 75-89
  14. 14.
    Grabska E, Hostert P, Pflugmacher D and Ostapowicz K. 2019. Forest stand species mapping using the Sentinel-2 time series. Remote Sensing, 11(10): 1197
  15. 15.
    Hościło A and Lewandowska A. 2019. Mapping forest type and tree species on a regional scale using multi-temporal Sentinel-2 data. Remote Sensing, 11(8): 929
  16. 16.
    Hu B X, Li Q and Hall G B. 2021. A decision-level fusion approach to tree species classification from multi-source remotely sensed data. ISPRS Open Journal of Photogrammetry and Remote Sensing, 1: 100002
  17. 17.
    Hu J, Shen L and Sun G. 2018. Squeeze-and-excitation networks//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE: 7132-7141
  18. 18.
    Immitzer M, Neuwirth M, Böck S, Brenner H, Vuolo F and Atzberger C. 2019. Optimal input features for tree species classification in Central Europe based on multi-temporal Sentinel-2 data. Remote Sensing, 11(22): 2599
  19. 19.
    Li S T, Kang X D, Fang L Y, Hu J W and Yin H T. 2017. Pixel-level image fusion: a survey of the state of the art. Information Fusion, 33: 100-112
  20. 20.
    Li X Y, Tian X, Duan T, Cao X M, Yang K J, Lu Q and Wang F. 2023. Estimation of fractional woody and herbaceous vegetation cover in Temperate Sparse Forest Grassland using fusion of UAV and Satellite imagery. National Remote Sensing Bulletin, 27(9): 2139-2152.
  21. 21.
    Liu Y N, Gong W S, Hu X Y and Gong J Y. 2018. Forest type identification with random forest using Sentinel-1A, Sentinel-2A, multi-temporal Landsat-8 and DEM data. Remote Sensing, 10(6): 946
  22. 22.
    Ma W L, Karakuş O and Rosin P L. 2022. AMM-FuseNet: attention-based multi-modal image fusion network for land cover mapping. Remote Sensing, 14(18): 4458
  23. 23.
    Mori A S, Lertzman K P and Gustafsson L. 2017. Biodiversity and ecosystem services in forest ecosystems: a research agenda for applied forest ecology. Journal of Applied Ecology, 54(1): 12-27
  24. 24.
    Persson M, Lindberg E and Reese H. 2018. Tree species classification with multi-temporal Sentinel-2 data. Remote Sensing, 10(11): 1794
  25. 25.
    Phiri D, Simwanda M, Salekin S, Nyirenda V R, Murayama Y and Ranagalage M. 2020. Sentinel-2 data for land cover/use mapping: a review. Remote Sensing, 12(14): 2291
  26. 26.
    Pimentel C S, McKenney J, Firmino P N, Calvão T and Ayres M P. 2020. Sublethal infection of different pine species by the pinewood nematode. Plant Pathology, 69(8): 1565-1573
  27. 27.
    Shafer G. 1992. Dempster-shafer theory. Encyclopedia of Artificial Intelligence, 1: 330-331
  28. 28.
    Shahdoosti H R and Ghassemian H. 2012. Spatial PCA as a new method for image fusion//The 16th CSI International Symposium on Artificial Intelligence and Signal Processing. Shiraz: IEEE: 90-94
  29. 29.
    Smith W B. 2002. Forest inventory and analysis: a national inventory and monitoring program. Environmental Pollution, 116: S233-S242
  30. 30.
    Snell J, Swersky K and Zemel R. 2017. Prototypical networks for few-shot learning//Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach: Curran Associates Inc.: 4080-4090
  31. 31.
    Tang C, Zheng X, Zhang W, Liu X W, Zhu X Z and Zhu E. 2023. Unsupervised feature selection via multiple graph fusion and feature weight learning. Science China Information Sciences, 66(5): 152101
  32. 32.
    Tian L, Tao Y, Fu W X, Li T, Ren F and Li M Y. 2022. Dynamic simulation of land use/cover change and assessment of forest ecosystem carbon storage under climate change scenarios in Guangdong Province, China. Remote Sensing, 14(10): 2330
  33. 33.
    Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez A N, Kaiser Ł and Polosukhin I. 2017. Attention is all you need//Proceedings of the 31st International Conference on Neural Information Processing Systems. Long Beach: Curran Associates Inc.: 6000-6010
  34. 34.
    Wang Z J, Chen B, Lu R Y, Zhang H, Liu H W and Varshney P K. 2020. FusionNet: an unsupervised convolutional variational network for hyperspectral and multispectral image fusion. IEEE Transactions on Image Processing, 29: 7565-7577
  35. 35.
    White J C, Coops N C, Wulder M A, Vastaranta M, Hilker T and Tompalski P. 2016. Remote sensing technologies for enhancing forest inventories: a review. Canadian Journal of Remote Sensing, 42(5): 619-641
  36. 36.
    Xia Q, Li J H, Dai S, Zhang H and Xing X M. 2023. Mapping high-resolution mangrove forests in China using GF-2 imagery under the tide. National Remote Sensing Bulletin， 27(6): 1320-1333.
  37. 37.
    Xu X D, Li W, Ran Q, Du Q, Gao L R and Zhang B. 2018. Multisource remote sensing data classification based on convolutional neural network. IEEE Transactions on Geoscience and Remote Sensing, 56(2): 937-949
  38. 38.
    Xu Y J, Zhou J and Zhang Z. 2024. A new Bayesian semi-supervised active learning framework for large-scale crop mapping using Sentinel-2 imagery. ISPRS Journal of Photogrammetry and Remote Sensing, 209: 17-34
  39. 39.
    Zhang L, Luo W T, Zhang H H, Yin X W and Li B. 2023. Classification scheme for mapping wetland herbaceous plant communities using time series Sentinel-1 and Sentinel-2 data. National Remote Sensing Bulletin, 27(6): 1362-1375
  40. 40.
    Zheng Y Z, Dong W, Yang Z P, Lu Y H, Zhang X, Dong Y N and Sun F Q. 2024. A new attention-based deep metric model for crop type mapping in complex agricultural landscapes using multisource remote sensing data. International Journal of Applied Earth Observation and Geoinformation, 134: 104204

Lire l'article complet

The above content is generated by Large Model Translation. The translated content is for reference only. We do not assume any commercial or legal responsibilty for any consequences arising from the use of our website