CNN-based transfer learning for forest aboveground biomass prediction from ALS point cloud tomography

• Verfasst von: Schäfer, Jannika [VerfasserIn]
• Verfasst von: Winiwarter, Lukas [VerfasserIn]
• Verfasst von: Weiser, Hannah [VerfasserIn]
• Verfasst von: Höfle, Bernhard [VerfasserIn]
• Verfasst von: Schmidtlein, Sebastian [VerfasserIn]
• Verfasst von: Novotný, Jan [VerfasserIn]
• Verfasst von: Krok, Grzegorz [VerfasserIn]
• Verfasst von: Stereńczak, Krzysztof [VerfasserIn]
• Verfasst von: Hollaus, Markus [VerfasserIn]
• Verfasst von: Fassnacht, Fabian Ewald [VerfasserIn]
• Titel: CNN-based transfer learning for forest aboveground biomass prediction from ALS point cloud tomography
• Verf.angabe: Jannika Schäfer, Lukas Winiwarter, Hannah Weiser, Bernhard Höfle, Sebastian Schmidtlein, Jan Novotný, Grzegorz Krok, Krzysztof Stereńczak, Markus Hollaus and Fabian Ewald Fassnacht
• E-Jahr: 2024
• Jahr: 08 Sep 2024
• Umfang: 18 S.
• Illustrationen: Illustrationen
• Fussnoten: Gesehen am 07.03.2025
• Titel Quelle: Enthalten in: European journal of remote sensing
• Ort Quelle: Florence : geoLAB, Laboratory of Geomatics, 2012
• Jahr Quelle: 2024
• Band/Heft Quelle: 57(2024), 1, Artikel-ID 2396932, Seite 1-18
• ISSN Quelle: 2279-7254
• DOI: doi:10.1080/22797254.2024.2396932
• Datenträger: Online-Ressource
• Sprache: eng
• Sach-SW: airborne laser scanning (ALS)
• Sach-SW: deep learning
• Sach-SW: Forest
• Sach-SW: random forest
• Sach-SW: synthetic data
• Sach-SW: virtual laser scanning
• K10plus-PPN: 1919292314

Abstract

This study presents a new approach for predicting forest aboveground biomass (AGB) from airborne laser scanning (ALS) data: AGB is predicted from sequences of images depicting vertical cross-sections through the ALS point clouds. A 3D version of the VGG16 convolutional neural network (CNN) with initial weights transferred from pre-training on the ImageNet dataset was used. The approach was tested on datasets from Canada, Poland, and the Czech Republic. To analyse the effect of training sample size on model performance, different-sized samples ranging from 10 to 375 ground plots were used. The CNNs were compared with random forest models (RFs) trained on point cloud metrics. At the maximum number of training samples, the difference in RMSE between observed and predicted AGB of CNNs and RFs ranged from −2 t/ha to 5 t/ha, and the difference in squared Pearson correlation coefficient ranged from −0.05 to 0.06. Additional pre-training on synthetic data derived from virtual laser scanning of simulated forest stands could only improve the prediction performance of the CNNs when only a few real training samples (10-40) were available. While 3D CNNs trained on cross-section images derived from real data showed promising results, RFs remain a competitive alternative.