Neural network has learned to identify tree species

Skoltech scientists have produced an algorithm that can identify a variety of tree species in satellite photos. Their research was posted in the IEEE Journal of Chosen Subject areas in Utilized Earth Observations and Distant Sensing.

Image credit score: Pixabay (Absolutely free Pixabay license)

Pinpointing tree species is crucial for successful forest administration and monitoring. Satellite imagery is an much easier and cheaper way to offer with this activity as when compared to other approaches that need floor observations of huge and distant locations.

Scientists from the Skoltech Centre for Computational and Facts-Intense Science and Engineering (CDISE) and Skoltech House Centre used a neural community to automate the identification of dominant tree species in significant and medium resolution photos. A hierarchical classification model and more data, these as vegetation height, assisted further more enrich the good quality of predictions when increasing the algorithm’s steadiness to aid its simple software.

Hierarchical model construction. Image credit score: EEE Journal of Chosen Subject areas in Utilized Earth Observations and Distant Sensing

“Commercial forest taxation providers and their conclusion end users, like timber procurers and processors, as well as the forest industry entities can use the new technological know-how for quantitative and qualitative assessment of wooden assets in leased locations. Also, our resolution permits quick evaluations of underdeveloped forest locations in phrases of investment decision attractiveness,” explains Svetlana Illarionova, the 1st author of the paper and a Skoltech PhD scholar.

There are plans to combine the produced algorithms in the Geoalert platform to automate the manufacturing of forest engineering resources promoted via Parma-GIS.

Source: Skoltech