What question did this study set out to answer?

To generate sub-meter resolution canopy height maps across large jurisdictions by combining high-resolution RGB satellite imagery with self-supervised deep learning models trained on aerial lidar data.

November 7, 2023Remote Sensing of EnvironmentOpen Access

Very high resolution canopy height maps from RGB imagery using self-supervised vision transformer and convolutional decoder trained on aerial lidar

View Full Paper

Ask AI

Bookmark

Authors

JTJamie Tolan HYHung-I YangUniversity of Minnesota BNBenjamin NosarzewskiCornell University

Discussion

Member takes

Implication

Computational modeling reveals sub-meter canopy height estimation from satellite RGB imagery across diverse forest ecosystems, indicating scalable individual-tree monitoring for carbon tracking.

Key Points

To generate sub-meter resolution canopy height maps across large jurisdictions by combining high-resolution RGB satellite imagery with self-supervised deep learning models trained on aerial lidar data.
Extracted high-level features from sub-meter Maxar RGB satellite imagery (2017–2020) using a self-supervised vision transformer model.
Trained a dense convolutional decoder against aerial lidar reference maps covering California and São Paulo, followed by post-processing with a convolutional network trained on spaceborne GEDI observations.
Evaluated model performance using set-aside validation lidar data, existing remotely sensed maps, and field-collected measurements.
Achieved an average Mean Absolute Error (MAE) of 2.8 m and a Mean Error (ME) of 0.6 m across validation aerial lidar datasets.
Delivered sub-meter canopy height maps across California and São Paulo, improving upon previous global 10 m resolution products to enable individual tree-level structural assessment.

Cite This Study

Tolan et al. (2023) studied this question.

synapsesocial.com/papers/69dc27eb45b398e6439f5334 https://doi.org/10.1016/j.rse.2023.113888

View Full Paper

Ask AI

Bookmark

Implication

Computational modeling reveals sub-meter canopy height estimation from satellite RGB imagery across diverse forest ecosystems, indicating scalable individual-tree monitoring for carbon tracking.

Very high resolution canopy height maps from RGB imagery using self-supervised vision transformer and convolutional decoder trained on aerial lidar

Discussion

Member takes

Implication

Key Points

Cite This Study

Very high resolution canopy height maps from RGB imagery using self-supervised vision transformer and convolutional decoder trained on aerial lidar

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider