Tree Inventory in Trollhättan

Organisation type

Public administration at local level

Use case area

Environment

Geographic scope

Sweden

Domain

Environment
Governance and implementation

Challenge:

The City of Trollhättan faces a significant challenge in managing its vast tree population across forest, park and urban environments. Manual inventory processes are resource-intensive, and there is often uncertainty regarding tree ownership within the municipality’s borders. Different owners, such as private individuals, municipal administrators, private businesses, government agencies, church communities and more, complicate the process. Additionally, identifying tree species is crucial for planning resilient green spaces.

Approach:

To address the challenge of managing its extensive tree population, the City of Trollhättan implemented a project to automate tree inventory processes. The Trollhättan AI-based tree inventory project utilised machine learning techniques to automate the analysis of the city's tree population. The team began by identifying and specifying the data requirements, followed by evaluating various data sources. They cleaned and processed the data to fit the chosen ML techniques. The project leveraged ortho-photos and LiDAR scans to create detailed 3D models of the city's green areas. AI models were trained to identify tree crowns and species, using both 2D and 3D data.
The AI model development involved several technical steps. The team gathered ortho-photos taken during different seasons and LiDAR point clouds to capture detailed geographical information. Manual annotation was performed to create training data, including georeferenced polygons for tree crowns. A Convolutional Neural Network (CNN) was used, requiring a unique architecture to integrate RGB imagery and LiDAR data. The model was trained to delineate tree crowns and classify tree species. The data was divided into training and test sets, with cross-validation to ensure model robustness. The project ensured data privacy and compliance with ethical guidelines, focusing on non-sensitive data.

Outcomes:

The project automates the tree inventory process, leading to more efficient and reliable data collection. Improved understanding and management of green areas will be achieved, contributing to sustainable urban planning. Decision-making will be enhanced regarding the city’s outdoor environments and green spaces. The project will also increase the city’s knowledge of data-driven methods, promoting machine learning-decision support. To achieve that, the team conducted workshops and seminars to enhance organisational learning and ensure the AI model's ethical and regulatory compliance. The final AI solution was handed over to Trollhättan's city officials with detailed documentation and training sessions to facilitate its integration into their urban planning processes.

Data Sources:

In this use case a variety of data sources are applied. Including orthophotos, which provide high-resolution aerial imagery. Lidar scanning data, offering precise 3D models of the terrain and existing geographic information which integrates various spatial data.

Additional Information:

Point of Contact:

kontaktcentertrollhattan [dot] se (kontaktcenter[at]trollhattan[dot]se)