How to conduct a forest inventory automatically with deep learning?

3D point cloud classification made smart, fast and easy.
Post Reply
User avatar
Francie Pointly
I have made 10-20 posts
I have made 10-20 posts
Posts: 11
Joined: Mon May 11, 2020 4:32 pm
Full Name: Francie Kastl
Company Details: Supper und Supper GmbH
Company Position Title: Business Development Manager
Country: Germany
Linkedin Profile: Yes
Location: Germany
Been thanked: 4 times

How to conduct a forest inventory automatically with deep learning?

Post by Francie Pointly »

This post is about a PoC we have recently done for RAG ( to automate a forest inventory by using AI.

A forest inventory can be a powerful management tool for forest management. Due to technological advances, it is possible to carry out a forest inventory using LiDAR scans and extract needed information automatically from point clouds. Recently, we developed an innovative, universal approach which enables identifying individual trees from different types of scans and gather relevant inventory data. If you want to know how we did it, this post is for you!

What is a forest inventory and its purpose?

An important part of forest management is an accurate forest inventory. A forest inventory is the gathering of information on the health and diversity of a community forest. It provides details such as tree condition, species, size and location.

A forest inventory may be conducted to:

- identify insect or disease problems of trees
- help the urban planners to develop maintenance and management plans
- analyze and increase the environmental impact of trees as a part of sustainable forest management
- determine the value of the community’s trees with regarding increasing property values
- inform the residents about their community forest
- visually assess timber and determine potential fire hazards and the risk of fire

Using LiDAR surveys for forest inventory

There are many methods how to conduct a forest inventory. In recent years both aerial and terrestrial LiDAR surveys have become widely used for this task as technological advances make it more affordable and precise. That is because more and more companies are developing cheaper LiDAR sensors and the availability of drones collecting LiDAR data is higher than ever. There has also been a big improvement in the quality of the scans as the sensors can emit more pulses per square meter and are able to collect data with extremely high point density. With the newest cloud solutions and high-performance computer processors, large amounts of 3D data can be stored, processed and organized.

„Recently, the use of unmanned aerial vehicles (UAVs) is quickly expanding and should revolutionize remote sensing in natural sciences.“ (source:

Nevertheless, extracting inventory information like a number of trees, their canopy shape or height is very labour-intensive as this is often done manually. A few automated methods and software solutions exist but they rely on very pronounced point cloud features like visible tree trunks or spatially separated trees.
A forest inventory conducted in a new way together with RAG


Part of our latest project for RAG was to conduct a forest inventory as a proof of concept. RAG specializes in long-term monitoring of mining damages, i.e. automatic detection of landscape changes, and here especially ground movements. A forest inventory can help to better isolate the causes of changes in the landscape through monitoring changes over time and observing which of them may be consequential damage to opencast mining.

The goal was to provide exact tree point identification in the point cloud as well as individual tree separation with corresponding inventory data:

- object-ID,
- position,
- crown’s extent

as a result.


LiDAR point clouds of trees can be extremely diverse depending on the applied acquisition method (airborne or terrestrial LiDAR scanning). That is why one of the biggest challenges was to develop algorithms based as little as possible on the specific characteristics of the different scan types, so that they could be used universally. Sometimes methods of recognizing tree trunk and treetops working on one point cloud were not suitable for another. Developing a universal approach required testing different methods like watershed separation and region growing from detected tree stems.

What difficulties did you face during either capturing forest scans or when trying to process them/work with them? Let me know in the comments :)


We combined deep learning and different smart algorithms to identify tree points and separate individual trees from above, both with visible tree trunks and without. Initially, the vegetation had to be classified manually to generate a comprehensive set of training data for the neural network so the process can be done automatically in the future. However, with the help of Pointly and its easy-to-use selection tools this task was performed in just a few days.

We first classified all trees and other objects with Pointly to proceed with instance segmentation of the individual trees. The segmentation was performed using a combination of tree features (such as crown maxima), tree stump positions and geospatial algorithms. Stump position and height were determined for each tree segment, whereas the crown extent was determined by means of 2D projection. Afterwards, the results (classification and instance segmentation) were transferred into other common geo data types (i.e. GeoJSON, Esri shape file), enriched with additional information such as height attributes.


Proof of concept outcome

The proof of concept in the project with RAG was successful. 2 map layers with tree counts, canopy shapes as well as tree height were generated. Now ongoing processing of new data is being performed so that this service can be automated and offered also directly in Pointly soon.
Pointly - Point out what matters
You do not have the required permissions to view the files attached to this post.
Best Regards
Francie Kastl
[email protected]
Post Reply

Return to “POINTLY”