Computer vision-based phone app to monitor forest health developed by University of Cambridge researchers

University of Cambridge researchers have developed an algorithm using computer vision techniques that will enable scientists to measure trees using their mobile phones.

The technique is four and half times faster than traditional, manual methods of ascertaining tree diameter - an important measurement used to monitor forest health and levels of carbon sequestration.

Measuring diameter at chest height tree by tree is time consuming, and prone to human error.

“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.”

The algorithm uses low-cost, low-resolution LiDAR sensors found in many mobile phones, and provides results that are equally as accurate than manual methods, the researchers report in the journal Remote Sensing.

While expensive, special-purpose LiDAR sensors can be used in forest measurements, they wanted to discover whether cheaper lower-resolution found in some phones for augmented reality applications could do the job.

This has only been tried before on highly-managed forests where trees are straight, evenly spaced and undergrowth is regularly cleared.

“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 Amelia.

The algorithm estimates tree diameter automatically from a single image in realistic field conditions and was incorporated into a custom-built app for an Android smartphones. It returns results in near real time.

The researchers first collected their own dataset by measuring trees manually and taking pictures. Then they used image processing and computer vision techniques to train the algorithm to differentiate trunks from large branches, determine which direction trees were leaning in, and other information to help it refine the information about forests.

They tested it in three forests – one each in the UK, US and Canada – in spring, summer and autumn. The app detected 100 per cent of tree trunks and had a mean error rate of eight per cent, which is comparable to the error rate when measuring by hand.

But it was far quicker.

“I was surprised the app works as well as it does,” said Amelia. “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 tool requires no specialised training so could be an accurate, low-cost tool for forest measurement, even in complex conditions.

The researchers plan to make their app publicly available for Android phones this spring.

The research was supported in part by the David Cheriton Graduate Scholarship, the Canadian National Research Council, and the Harding Distinguished Postgraduate Scholarship.