Watershed vs. Region Growing for Individual Tree Segmentation from Airborne LiDAR: An Urban Case Study in Bologna
This study compares watershed and region growing algorithms for segmenting individual urban trees from airborne LiDAR in Bologna, revealing distinct detection profiles for each method while highlighting the limitations of existing municipal records as ground truth and demonstrating a modular pipeline for deriving vegetation indicators like biomass and pollen production.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are a detective trying to solve a mystery in a giant, invisible forest. But this isn't a forest of dirt and leaves you can touch; it's a forest of light beams. Scientists use a special tool called LiDAR (Light Detection and Ranging) to shoot lasers from the sky. These lasers bounce off everything—buildings, cars, and trees—and come back to tell us exactly how high they are. It's like the city is wearing a 3D suit made of millions of tiny dots.
Now, imagine you have a massive pile of these dots, and your job is to figure out which dots belong to which tree. This is called segmentation. It's tricky because trees in a city often grow close together, their branches tangle like a bowl of spaghetti, and some are tall while others are short. If you can't tell one tree from another, you can't count them, measure their health, or guess how much carbon they are soaking up from the air. This matters because cities are getting hotter, and trees are nature's air conditioners. To manage them, we need to know exactly who is who in the green crowd.
This paper is about two different detectives trying to solve this puzzle in Bologna, Italy. The first detective uses a method called Watershed, which is like pouring water over a bumpy landscape. The water flows down into valleys, and where the water from two different valleys meets, that's the border between two trees. The second detective uses Region Growing, which is more like a game of "connect the dots." It starts at the very top of a tree and grabs onto nearby points that look like they belong to the same family, growing the tree outward until it hits a neighbor.
The researchers tested both detectives on 12 different slices of the city. Here is what they found: The Watershed detective found more trees—7,589 of them! But, it seemed a bit too eager. It chopped up big trees into many small pieces, finding a huge number of tiny trees that were only about 5 meters tall. It was like seeing a giant oak tree and thinking it was actually a forest of small saplings because the "water" got stuck in the little dips of the leaves.
The Region Growing detective was more conservative. It found fewer trees—6,432—but it was better at spotting the giants. It successfully identified trees that were over 40 meters tall, which the other method missed. This happened because Region Growing didn't smooth out the data; it kept the sharp, high points of the tallest trees intact.
The team also tried to check their work against the city's official list of trees, a database called Alberi in manutenzione. They hoped to use this list as a "truth" to see who was right. But the list turned out to be a bit of a mess. About half the trees the LiDAR saw weren't on the list at all. Some spots on the list had trees that were no longer there, and the height records were mostly from 20 years ago. It was like trying to check your homework against a textbook from a decade ago that had half the pages missing. Because of this, the authors couldn't say for sure which detective was perfect, but they could see that the city's old list wasn't good enough to be the referee.
Despite the confusion, the team used their new tree maps to calculate some cool numbers. They estimated how much carbon each tree stores (which helps fight climate change) and how much pollen they might produce (which can make people sneeze). They turned all this data into an interactive map where you can click on a tree and see its height, its size, and its "pollen power."
The big takeaway? The city's old tree list is outdated and unreliable. The new LiDAR methods work, but they need to be tuned carefully. The "Watershed" method is fast but might see too many small trees, while "Region Growing" is great for spotting the giants but might miss some details. The authors suggest that in the future, we should combine these sky-scans with ground-level scans and fresh field surveys to get the perfect picture. Until then, this new pipeline gives the city a powerful, up-to-date way to see its greenery, even if the old maps are still playing catch-up.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.