Neural network estimates forest height using satellite imagery
Image credit: Lane Erickson/Dreamstime
Researchers have trained a neural model to determine the height of trees to monitor the natural environment, infrastructure, and timber supply.
A neural network could size up trees from satellite images, according to new research.
From environmental scientists to civil engineers and wood industry workers, there are many people who require accurate estimates of forest tree size. This information is vital for understanding how much atmospheric carbon dioxide the trees are capturing, whether there’s risk of them damaging power lines, and how much timber is available for logging.
Currently, these estimates are obtained from satellite images as well as multiple cameras spanning several bands of infrared radiation, as drone technology is ineffective in large and hard to reach regions. However, this multispectral data is scarce and expensive to acquire.
Researchers from Moscow-based research institute Skoltech may have found the perfect alternative, as they have been able to train a neural model to determine tree height in a reliable and cost-effective manner.
Unlike prior solutions, the model presented in IEEE Access does not require drone footage or imaging beyond the visible range. Instead, the neural networks only uses ordinary optical satellite imagery,
“The one biggest factor that makes our neural network successful is its ability to analyse spatial data and texture characteristics," said Skoltech PhD student Svetlana Illarionova.
"Along with the optical imagery, we put in supplementary features in the form of ArcticDEM, a freely available high-resolution model. It is a 2-metre-resolved representation of the bare topographic surface of the Earth covering boreal regions.”
The researchers have exploited the connection that exists between tree crown shape and height to create the high-quality model.
Training data comes from the northern Russian region of Arkhangelsk, and the canopy height predictions are scored based on how well they match lidar observations made on location in that region with drones. The researchers say their solution is applicable to any sites with similar vegetation.
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.