GEMMA FUSION

GeMMA Fusion Suite is advanced geospatial software SDK and tool designed and optimized for mapping of geospatial shape objects from heterogeneous sensory data streams and remote sensing data sources, such as digital ortophoto, lidar, digital terrain model and satellite imagery. GFS core consists of optimized readers and writers for standard data formats that allow for efficient async data IO operations and large-scale processing. Initiated by a growing number of studies conducted by GeMMA Lab with its industrial partners, GFS analytic engine is being constantly enriched by advanced data fusion methodologies and new location-intelligence algorithms. Over the years, codebase evolved into fully-fledged high-performance SDK to efficiently solve any given geoenvironmental and spatial data fusion problem.

Development of GFS was confunded by:

As sown in figure above, GFS consist of the following system components:

• GF Spatial is a collection of APIs for spatial data management and processing that allows for consuming or authoring various spatial data products and maps in form of vector, raster, poincloud and fullmotion video files or services.
• GF Processing is collection on APIs for feature processing and extraction from given spatial data. It consists of powerful algorithms for raster, geometry and pointcloud processing and methods for conversion between data types such as vectorization of raster objects or poinclouds and 3D rasterization of geometry objects or pointclouds.
• GF Rendering and UI is set of high performance APIs for efficient spatial data rendering using GPU and various UI widgets that allow for data editing which are easily included in other native applications.
• GFS Mapping Tool is a desktop application that allows for efficient generation of various data products and maps. In addition to all fundamental tools for processing of raster, vector, and attribute data, it offers a wide range of AI methods for automatic processing of heterogenous data layers, including IoT data and video streams that can be used for (near) real-time map generation and distribution.
• GFS Automated Learning Tool is a desktop application, currently under development. When implemented, it will allow users to easily generate learning data, train AI models on it, and integrate trained models into GFS Mapping Tool for their everyday use.
• GFS Data Space is a back-end data distribution system that allows for distribution of their own data products, as generated by GFS Mapping tool, integration of third-party data distribution system and IoT systems, and harvesting the data from third party data providers.
• GFS Fog AI is system intended for management of machine learning models that is currently under the development. The system will offer a full stack of tools for storing, administering, and distributing AI models, with the support of transfer and federated learning services.

Gemma Fusion Suite contains many basic as well as advanced features which are easy to use. Most important feature is ability to load large quantities of multimodal data such as rasters, vectors and point clouds and group them into the workspaces (so called projects).

Supported data, databases and services:

• Shapefile
• Geojson
• Geopackage
• Tiff Images
• Jpeg2000
• Postgres
• MSSQL
• WMS, WMTS
• TMS
• WCS
• ArcGIS HTTP Services

Functionality

• Digitalization: Create new shape features as fast as possible without any workflow interuptions.
• Layers and attributes: Manage spatial layers, styles and attribute schemas. Edit feature properties and define automatic rules.
• Map services: Import data from standard online services such as TMS, WMS, WMTS, WCS, ArcGIS HTTP Services
• Lidar classification: Automatically classify imported lidar data. GFS also supports  intuitive manual point cloud classification.
• Advanced geometry tools: Move shape/points or add/remove points on the existing geometry. Perform union, difference or intersection on the selected shapes and many more.
• Raster vectorization: Extract raster features as shape features and attribute them afterwards
• 3D geometry rasterization: Rasterize 3D features onto rasters.

In comparison to the competitors, GFS provides:

• Integration of heterogeneous data sources in addition to basic raster and vector data, like LiDAR, IoT, and video streams,
• Map editing tools, namely general tools used for editing raster or vector data layers and associated attribute data, like drawing vector object, adding new attributes or removing old ones or performing basic image processing, as for example edge detection, noise removal, object segmentation,
• Predefined AI models for automatic object recognition, as for example recognition of vegetation in LiDAR data or building recognition in satellite data,
• (Near) real-time AI support addresses integration of real-time data streams like IoT data or video streams for rapid generation of maps,
• Data fusion services which enable feature extraction from multiple complementary data sources for improved object and situation assessment as for example integration of satellite and LiDAR data for predictions of vegetation growth.
• Automated learning, where users can produce their own AI models based on their own learning data and integrate it into a mapping tool,
• Federated AI Support that allows for management of models, their additional training and specialisation, as for example using general vehicle recognition models to train car recognition models as well as combining different models into one model, as for example building recognition model trained on Chinese architecture with building recognition models trained on Central European architecture.
• Fog AI support with dedicated tools for distributing the models to edge devices, like for example mobile devices and embedded systems.
• Distribution of data products where users can submit the newly generated data products on the distribution system to be immediately used by their costumers all around the globe.

Following is the list of the existing users, with a short description of the means of their usage:

• Slovenian Environment Agency used GFS for mapping all water streams in Slovenia using fusion of LiDAR data and Digital Orthophoto,
• GeoDirectory used GFS for enriching their maps of addresses in Ireland with the solar potential of associated rooftops by a fusion of LiDAR and satellite data,
• Petrol d.d. uses GFS for measuring water levels in accumulation lakes and consequential estimation of hydroelectricity production potentials using fusion of LiDAR generated DTMs and Sentinel 2 data,
• ELES d.o.o. for vegetation management in powerline corridors that includes vegetation growth simulation based on the fusion of LiDAR data, thematic maps about soil quality and distribution of tree species, as well as weather data,
• DARS d.o.o. uses GFS for short- and long-term traffic predictions using real-time fusion of traffic counters and floating card data with weather data and contextual data, e. g. about holidays in the neighbouring countries.
• Ministry of Defence used GFS for real-time fusion of with thematic maps and video-streams from drones, with detection of vehicles, buildings, and people for rapid mapping of battlefields.
• Chamber of Agriculture and Forestry of Slovenia uses GFS for fusion of relevant environmental data sources about vegetation growth that feeds their system of farm economy calculations and planning crop cycles.
• DEWESoft d.o.o. uses parts of GF SDK for raster/terrain map processing and serving to their clients via MapVC widget. Engine for 3D map rendering is powered by GF SDK.

Potential users thus include individuals, companies, and public institutions that need to develop improved understanding of the environment for their business processes.

Source code, precompiled SDK and application releases can be found under internal GitLab repository, which is only available for the registered users bellow:

GemmaFusion Releases

For additional information, contact us via email at

gemma.feri@um.si