How do we measure a tree? In most cases, researchers use manual techniques on the ground, measuring the diameter at chest height. While this method is reliable, it can also be time-consuming and risks human error. Now, a team at the University of Cambridge has developed an algorithm to measure trees with our smartphones.
Measuring trees is a common task for some researchers -- and a very important one. It provides relevant information about the health of trees and the wider forest ecosystem, as well as how much carbon is being sequestered. When trees perform photosynthesis, they take carbon dioxide out of the air, bind it into sugar, and then release oxygen.
“When you’re trying to figure out how much carbon a forest is sequestering, these ground-based measurements are hugely valuable, but also time-consuming,” first author Amelia Holcomb from Cambridge University’s Computer Science department said in a statement. “We wanted to know whether we could automate this process.”
A different way to measure trees
Holcomb and her team developed an algorithm that gives an accurate measurement of tree diameter, using LiDAR sensors that are already incorporated into many mobile phones. This gives results that are as accurate as manual measurement techniques but much faster, the researchers said. The results are reported in a study in the journal Remote Sensing.
As a reminder, LiDAR stands for light detection and ranging. It uses a laser to measure distances and map objects. In the past, LiDAR systems were very expensive and bulky, but with the advancements in technology, the sensors are now in most phones. They are used, for example, to take better photos by adding depth scanning techniques.
Other researchers have done forest measurement studies using LiDAR sensors. However, they focused on managed forests where trees are straight, evenly spaced and undergrowth is cleared. Holcomb wanted to test whether these sensors could give accurate results for non-managed forests quickly, automatically, and in a single image.
They designed an algorithm that estimates trunk diameter automatically from a single image in realistic field conditions. The algorithm was then incorporated into an app for an Android smartphone that is able to return results in near real-time. The app isn’t publicly available yet, but the researchers hope to make it available on Android devices later this spring.
To create the algorithm, they created their own dataset by measuring trees manually and taking photos. Using image processing and computer vision techniques, they trained the algorithm to differentiate trunks from large branches and determine which direction trees were leaning in. Then it was time to test it, which they did in three forests.
They used their app in forests in the UK, the US and Canada in spring, summer and autumn, and overall it was a big success. It was able to detect 100% of tree trunks and had an average error rate of 8% - which is similar to the error rate when measuring trees manually. The app was also four and a half times faster than measuring by hand.
The researchers were surprised by how well the app works. It requires no specialized training to use and it builds on the sensors that are already incorporated on many phones. This means it could eventually become an accurate and low-cost tool to measure forests even in complex forest conditions, they said.