Science & Technology

Neural networks have learned to identify tree species from satellites

Detailed land cover showing forests in Chiapas, southern Mexico. The map was created using Copernicus Sentinel-2 optical data from April 14, 2016. Images are not part of the study discussed.

Much of what we know about forest management comes from recent aerial photography. Bird’s-eye views of forests, whether drones, helicopters, or satellites, are important for understanding how forests are moving, especially in remote areas that are difficult to monitor on the ground.

Satellite imagery in particular provides an inexpensive and effective surveillance tool. However, the problem with satellite data is that they are often quite low resolution and it can be difficult to tell what you are looking at.

However New research It can be helpful to use neural networks to distinguish satellite images.

Hierarchical model structure / Svetlana Illlarionova et al. , IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

“Commercial forest tax providers and their end users (including timber suppliers and processors), as well as forestry industry associations, can use new technologies for the quantitative and qualitative assessment of timber resources in leased areas. Our solution also enables a rapid assessment of undeveloped forest areas in terms of investment appeal, ”explains Svetlana Illlarionova, lead author of the paper and Skoltech PhD student.

Illlarionova and her colleagues at the Skoltech Center for Computational and Data-Intensive Science and Engineering (CDISE) and Skoltech Space Center used neural networks to automate the identification of dominant tree species in high- and medium-resolution images. ..

Class markup for the study area. Image credit: Illlarionova et al.

After training, the neural network was able to identify the predominant tree species at a test site in Leningrad, Russia. The data was confirmed by ground observations in 2018. Additional data, such as the stratification model and vegetation height, helped to further improve the quality of the predictions while improving the stability of the algorithm and facilitating its practical application.

This study focused on identifying the dominant species. Of course, some forests with different compositions have approximately the same distribution of two or more species, but the composition of these mixed forests was outside the scope of the study.

“It is worth noting that the” dominant species “in forestry does not exactly match the biological term” species “and is primarily related to the type and quality of timber,” the researchers said. I am writing.

Overall, the algorithm seemed to be able to identify the predominant species, but researchers say that better training markups planned for future research can improve results.

“But future studies will cover the case of mixed forests, which fall into a fully hierarchical segmentation scheme. Another goal is to add forest inventory characteristics that can also be estimated from satellite images. The study concludes.

Neural networks have learned to identify tree species from satellites Neural networks have learned to identify tree species from satellites

Back to top button