Researchers from the University of Cambridge, UK, developed an algorithm that uses smartphone sensors to estimate trunk diameter automatically from a single image in realistic field conditions, according to a study published in the journal Remote Sensing. It is five times faster than the traditional method and can be included in an app for Android phones.
The standard measurement in ecology is the tree diameter at chest height. This data can be used to determine how healthy specific trees are, as well as assess the wider ecosystem and calculate how much carbon is sequestered in a certain area.
The problem is that this can be a time-consuming process. “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,” said first author Amelia Holcomb from Cambridge’s Department of Computer Science and Technology. “We wanted to know whether we could automate this process.”
To speed up the measurements, the team developed a new algorithm using low-cost, low-resolution sensors that can be included in mobile phones to produce results that are just as accurate as more expensive equipment or manual measurements.
This is not the first time these sensors have been used for forest measurements. However, previous attempts measured highly-managed forests with perfectly straight trees, evenly spaced out, and limited undergrowth. The Cambridge team wanted to test whether the sensor could be used in less controlled conditions and still produce accurate results. “We wanted to develop an algorithm that could be used in more natural forests, and that could deal with things like low-hanging branches or trees with natural irregularities,” said Holcomb.
The first step was to develop the algorithm. For this, the team collected their dataset by manually measuring trees and taking pictures. Then, using image processing tools, they “trained” the algorithm to separate between large brunches and trunks, as well as determine which direction trees were leaning.
Once it was ready, the authors tested the algorithm in three different forests (in the UK, US, and Canada) in the spring and repeated it again in the fall. The app detected 100% of tree trunks with an error rate of 8%, which is similar to error rates with hand measurements. However, compared to manual measurements, the app allowed researchers to be five times faster. “I was surprised the app works as well as it does,” said Holcomb. “Sometimes I like to challenge it with a particularly crowded bit of forest or a particularly oddly-shaped tree, and I think there’s no way it will get it right, but it does.”
The authors believe this could become an accurate and low-cost tool for forest measurements, even in complex conditions. The researchers now want to make their app available for Android phones later this spring.
Holcomb, A.; Tong, L.; Keshav, S. Robust Single-Image Tree Diameter Estimation with Mobile Phones. Remote Sens. 2023, 15, 772. https://doi.org/10.3390/rs15030772