èƵ

Google Street View helps map how 600,000 trees grow down to the limb

AI and Google Street View have created 'digital twins' of living trees in North American cities – part of a huge simulation that could help make urban tree planting and trimming decisions
Predicting how trees grow could help guide planting efforts in urban areas such as New York City
Yukinori Hasumi/Getty Images

Artificial intelligence coupled with Google Street View images has created hundreds of thousands of “digital twins” of trees across North America. The simulation could help city planners better predict how seasonal foliage boosts cooling or when growing branches may require trimming.

“If you can model the existing set of trees and you have a reasonable model of infrastructure like power lines, you can understand where you’re going to have trees growing into power lines that are potentially most harmful,” says at the Massachusetts Institute of Technology.

Beery and her team used AI to transform 2D photos taken from Google Street View into 3D “digital twins” of 600,000 living trees across 23 North American cities. They trained the AI model on labelled images of trees from over 300 genera, from pine to maple, and on separate 3D tree models generated by another computer algorithm.

The trained AI then successfully produced 3D models of the trees captured in Google Street View images, including predicting how they would grow even accounting for other nearby trees or buildings.

This AI tool works best in digitally recreating common North American trees but can struggle to identify palm trees, for example. That is because the tree’s fronds are truncated in the street-level images, says Beery. The 3D modelling also has difficulty identifying separate trees with entangled branches.

Living tree growth simulated in 3D using Google Street View images
Jae Joong Lee

The Google Street View images that this method relies upon do not capture every tree hidden in backyards or parks, says at Canadian-based telecommunications company Telus. Still, she says the new tool could reduce the need for human inspections and help improve tree planning decisions. “Urban forests are critical to urban resilience and livability.”

This AI-powered tree simulation could someday help decide what trees to plant on a future warmer planet, as the regions where these trees grow could shift thousands of miles within the next 20 to 50 years, says Beery. “Trees that are happiest growing in North Carolina now will be happiest growing in New York City in just a few decades,” she says.

Reference:

arXiv

Topics: Artificial intelligence / Climate / Trees