Skip to content

Welcome to AImon5.0

AImon5.0 logo 3DGeo Logo

License: MIT

Main 3DGeo outcomes of the AImon5.0 project

📑 Research project at a glance

Modern permanently installed laser scanners can deliver sub‐hourly point clouds, opening the door to early warning of surface deformations. Current workflows struggle to keep pace with such data volumes in near real-time. Vegetation and occlusions in forested or complex terrain further degrade ground‐point coverage, undermining the reliability of change estimates. As a result, there remains a critical need for efficient, robust processing strategies that can detect and quantify subtle surface shifts.

The project AImon5.0 was funded by the Federal Ministry of Research, Technology and Space (BMBF) within the funding measure "Digital GreenTech – Environmental Engineering meets Digitalisation" as part of the "Research for Sustainability (FONA) Strategy". Please find more project details on our website.

Overall objective

Our environment and the Earth's surface are constantly changing, and global warming and climate change are accelerating the pace and magnitude of these changes. As a result, geohazards, triggered by natural events or human activities, are becoming more frequent. For example, intense and prolonged rainfall or thawing of permafrost in the Alps are increasingly causing landslides and rockfalls that threaten local populations and critical infrastructure such as railways and roads, with serious economic consequences.

A key tool for integrated risk management is access to relevant 4D geospatial information (accurate 3D data with high temporal resolution) acquired through near real-time, permanent or on-demand monitoring. Permanently installed autonomous laser scanning (PLS) systems have shown great potential for monitoring hazard zones, producing billions of 3D measurements daily. Computational methods to analyze this data exist, but to make it operational, a new interface is needed to bridge application needs with 4D data collection and analysis.

This interface connects stakeholder expertise with autonomous PLS systems and data archives using AI and 4D analysis. It enables the operational use of PLS for risk monitoring - detecting and tracking relevant events such as slope activity in real time. For the first time, stakeholders are able to use PLS for continuous hazard monitoring.

Study site

The goal of this project was to bridge the gap between research and practice. While key methods for multi-temporal analysis and subtopics such as uncertainty in change detection had been previously developed, this project focused on refining and extending them for practical, application-oriented use. The 3DGeo Research Group Heidelberg developed computer-based methods for automatic information extraction and visualization from 4D-PLS data as their main focus in the AImon5.0 project. The study site was located in Trier, Germany (Fig. 1), at the Trierer Augenscheiner (Fig. 2).
Trier Map
Study site of the AImon5.0 project located in Trier in Germany (red dot).

Method development and implementation

The developed methods are particularly suitable for operational use and adapted in order to deliver reliable and timely results. Automated information extraction represents a central interface between the PLS system in the field, the quality-assured change information, and the end users. With georeferenced point clouds using the open-source method Piecewise-ICP, we scientifically investigated and combined two complementary concepts that can integrate expert knowledge into automated data analysis:
     1. Top-down approach via a knowledge- and rule-based classification of changes: In that case, the users know exactly which events they want to find in the data streams and how these processes (e.g., abrupt rockfall) are defined in their sequence. A set of new methods and tools for data management for fast and accurate searches were developed and evaluated;
     2. Data-driven approach using AI: Machine learning methods find relevant change events after a user-controlled training phase and present them to the experts for evaluation. The users do not know in advance how the events, possibly also overlaid processes, are represented in the raw data. However, the users can distinguish relevant from non-relevant extracted events for their use case and thus train an AI model.
For the second approach, research was carried out to find out how the state-of-the-art point cloud-based machine learning models can be trained quickly and as automatically as possible in the background and how their hyperparameters can be optimized. In coordination with end users and the PLS operator, it was determined which abstraction levels and visualization forms are best suited for certain tasks, as well as specified reaction times, for the visualization of the detected and classified changes. In contrast to visualization in 2D and 3D (e.g., in GIS or dashboards), fundamental research was carried out for the visualization of 4D processes in PLS data due to a lack of existing methods and tools.

⚙️ Work packages

The following methods were developed by the 3DGeo Research Group within their main work packages.

New concept of 'change events'

A change event is characterized by different attributes, each representing a measurable dimension of the change. The diagram of Fig. 3 highlights the modular structure of change events, emphasizing how temporal and spatial metrics combine to define and categorize observed changes. A classifier assigns an event type to the change based on the characteristics, enabling semantic interpretation of what kind of event occurred (e.g., gravitational mass movement, change in vegetation, etc).
logo

Research target 1 - Hierarchical classification of detected change:

Developing new methods and tools to automatically extract relevant change information from the last two point clouds. The method analyses different types of change in the terrain (e.g., rockfall events, movements or erosion processes) fully automatically by delimiting them in terms of time and space. Five different steps comprise:
  • 1.1: Rule-based change classification
  • 1.2: Hierarchical analysis
  • 1.3: ML/DL change classification
  • 1.4: Derivation of adaptive workflows
  • 1.5: Continuous integration in py4dgeo

Research target 2: Visualization of classified change events:

Development of new concepts and tools for the visualization of the detected terrain changes for use by end users. Three different steps comprise:
  • 2.1: Selection of relevant changes
  • 2.2: 2D GIS layer
  • 2.3: 3D objects

💡 Implemented methods and their potential applications

  • Point cloud projection: Generate range and color images from point cloud data.
  • Bi-temporal analysis: Compare point clouds from different time frames to detect changes.
  • Change event management: Convert detected clusters into change events.
  • Data handling: Efficiently split, append, and merge LAS/LAZ files.
  • 3D objects: Convert change events into 3D mesh objects.
  • GIS and KML layer generation: Project 3D change events to 2D GIS and KML polygon layers with their metadata for QGIS and Google Earth visualization.
  • Visualization: Quickly visualize change events on the generated range and color images.

🎮 Examples

Example 1: Main AImon5.0 monitoring pipeline

Example 2: Adaptive monitoring
Example 3: Rule-based classification of change events
Example 4: Rule-based filtering of change events
Example 5: Manually labelled dataset for random forest training
Example 6: Random forest classification on prediction dataset

🛠️ Full worfklow implementation

Serves as the entry point for the AImon5.0 processing workflow. It orchestrates the execution of various processing stages, including configuration setup, bi-temporal analysis, and change detection.

📚 Publications

Journal and conference 
@article{Tabernig2025,
  author       = {Tabernig, Ronald and Albert, William and Weiser, Hannah and H{\"{o}}fle, Bernhard},
  journal      = {6th Joint International Symposium on Deformation Monitoring (JISDM)},
  year         = {2025},
  title        = {A hierarchical approach for near real-time 3D surface change analysis of permanent laser scanning point clouds},
  doi          = {10.5445/IR/1000180377},
  pagetotal    = {9}
}

@article{WeiserHoefle2025,
  author       = {Weiser, Hannah and H{\"{o}}fle, Bernhard},
  journal      = {EarthArXiv},
  year         = {2025},
  title        = {Advancing vegetation monitoring with virtual laser scanning of dynamic scenes (VLS-4D): Opportunities, implementations and future perspectives},
  doi          = {10.31223/X51Q5V},
}

@article{Albert2025,
  author       = {Albert, William and Weiser, Hannah and Tabernig, Ronald and H{\"{o}}fle, Bernhard},
  journal      = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
  year         = {2025},
  title        = {Wind during terrestrial laser scanning of trees: Simulation-based assessment of effects on point cloud features and leaf-wood classification},
  doi          = {},
  pagetotal    = {8}
}

@article{Meyer2025,
  author       = {Meyer, Jannik and Tabernig, Ronald and H{\"{o}}fle, Bernhard},
  journal      = {ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
  year         = {2025},
  title        = {Detection of honey bees (Apis mellifera) in hypertemporal LiDAR point cloud time series to extract bee activity zones and times},
  doi          = {},
  pagetotal    = {8}
}
Software 
@software{Tabernig_VAPC_-_Voxel_2024,
  author = {Tabernig, Ronald and Albert, William and Weiser, Hannah and H{\"{o}}fle, Bernhard},
  license = {MIT},
  month = dec,
  title = {{VAPC - Voxel Analysis for Point Clouds}},
  url = {https://github.com/3dgeo-heidelberg/vapc},
  version = {0.0.1},
  year = {2024}
}
Abstract 
@inproceedings{TabernigAGIT2025,
  author       = {Tabernig, R. and Albert, W. and Weiser, H. and H{\"{o}}fle, B.},
  title        = {Towards in-situ near real-time 3D environmental monitoring and geospatial point cloud analysis with open-source software},
  booktitle    = {AGIT Conference 2025. Vol. 1},
  editor       = {},
  pages        = {202},
  year         = {2025},
  address      = {Salzburg},
  publisher    = {Universitätsbibliothek Salzburg},
  doi          = {10.25598/agit/2025-48},
  note         = {Abstract},
  url          = {https://doi.org/10.25598/agit/2025-48}
}

@inproceedings{AlbertSummerSchool2024,
  author       = {Albert, William},
  title        = {Considering wind effects in LiDAR simulation-based machine learning for point cloud classification in forests},
  booktitle    = {Sensing Mountains. Innsbruck Summer School of Alpine Research 2024 – Close Range Sensing Techniques in Alpine Terrain},
  editor       = {Rutzinger, M. and Anders, K. and Eltner, A. and Gaevert, C. and Höfle, B. and Lindenbergh, R. and Mayr, A. and Nopens, S. and Oude Elberink, S. and Pirotti, F.},
  pages        = {19--22},
  year         = {2024},
  address      = {Innsbruck},
  publisher    = {innsbruck university press (IUP)},
  doi          = {10.15203/99106-137-3},
  note         = {Abstract},
  url          = {https://doi.org/10.15203/99106-137-3}
}

@inproceedings{Hofle2024,
  author       = {H{\"{o}}fle, B. and Tabernig, R. and Zahs, V. and Esmorís, A.M. and Winiwarter, L. and Weiser, H.},
  title        = {Machine-learning based 3D point cloud classification and multitemporal change analysis with simulated laser scanning data using open source scientific software},
  booktitle    = {EGU General Assembly 2024},
  volume       = {EGU24},
  pages        = {1--2},
  year         = {2024},
  doi          = {10.5194/egusphere-egu24-1261},
  note         = {Abstract},
  url          = {https://doi.org/10.5194/egusphere-egu24-1261}
}


@inproceedings{TabernigSummerSchool2024,
  author       = {Tabernig, Ronald},
  title        = {Simulating laser scanning of dynamic virtual 3D scenes for improved 4D point cloud based topographic change analysis},
  booktitle    = {Sensing Mountains. Innsbruck Summer School of Alpine Research 2024 – Close Range Sensing Techniques in Alpine Terrain},
  editor       = {Rutzinger, M. and Anders, K. and Eltner, A. and Gaevert, C. and Höfle, B. and Lindenbergh, R. and Mayr, A. and Nopens, S. and Oude Elberink, S. and Pirotti, F.},
  pages        = {134--137},
  year         = {2024},
  address      = {Innsbruck},
  publisher    = {innsbruck university press (IUP)},
  doi          = {10.15203/99106-137-3},
  note         = {Abstract},
  url          = {https://doi.org/10.15203/99106-137-3}
}

@inproceedings{Tabernig2024,
  author       = {Tabernig, Ronald and Zahs, Vivien and Weiser, Hannah and H{\"{o}}fle, Bernhard},
  title        = {Simulating 4D scenes of rockfall and landslide activity for improved 3D point cloud-based change detection using machine learning},
  booktitle    = {EGU General Assembly 2024},
  year         = {2024},
  address      = {Vienna, Austria},
  month        = apr,
  note         = {EGU24-1613},
  doi          = {10.5194/egusphere-egu24-1613},
  url          = {https://doi.org/10.5194/egusphere-egu24-1613}
}
Bachelor thesis 
Lukas Fuchs (2025): Assessment of Low-Cost Laser Scanning Setups for 3D Rockfall Monitoring Using Virtual Laser Scanning. Institute of Geography, Heidelberg University.

Niklas Carniel (2024): Deriving activity zones of the Äußeres Hochebenkar rock glacier through boulder-tracking based on multitemporal ULS point clouds. Institute of Geography, Heidelberg University.

Michelle Meier (2024): Change Detection in Laser Scanning Data for Rockfall Classification and Correlation Analysis on a Slope in Obergurgl, Tyrol. Institute of Geography, Heidelberg University.

📂 Credits

Software usage

As a starting point, please have a look to the Jupyter Notebooks available listed in the top left corner of the page. Click here for a detailled description of the configuration file parameters.

Citation

Please cite the AImon5.0 repository when using our software & tools in your research.

@software{AImon5.0,
author = {Albert, William and Tabernig, Ronald and H{\"{o}}fle, Bernhard},
title = {AImon5.0: tool for 3D point cloud processing and projection},
journal = {},
year = {2025},
number = {},
volume = {},
doi = {},
url = {https://github.com/3dgeo-heidelberg/AImon},
}

Funding / acknowledgements

The Federal Ministry of Research, Technology and Space (BMBF) was funding the AImon5.0 project within the funding measure "Digital GreenTech – Environmental Engineering meets Digitalisation" as part of the "Research for Sustainability (FONA) Strategy".

Contact / bugs / feature requests

Have you found a bug or have specific request for a new feature? Please open a new issue in the online code repository on Github. Also for general questions please use the issue system. Scientific requests can be directed to the 3DGeo Research Group Heidelberg and its respective members.

License

This is licensed under the MIT license.