For more than 20 years, a dedicated team of engineers and researchers has conceived innovative solutions, at the cutting edge of AI and photogrammetry technologies, to meet the needs of all the group and our customers.
Our innovation is mainly focused on topics related to:
- Image analysis and processing.
- Photogrammetry.
- LiDAR point clouds analysis.
- Geospatial data production.
Our achievements
Earth Observation
Our innovation team has developed several AI image analysis solutions for Earth observation. These images have various sources, from satellites such as Pléiades and Pléiades NEO to multispectral aerial images. Implemented tools are based on Machine Learning and Deep Learning and aim to detect specific types of objects (building, …) or to segment i.e. to assign a class (crop field, road, …) to each pixel.
Data: Sentinel 2 images, Pléiades images, Pléiades NEO images.
Urbanism
Our innovation team has developed its own Deep Learning model for the automatic segmentation of point clouds based on a state-of-the-art model, optimized and adapted to our industrial constraints. This model is used to support street plans production.
In the same domain, Innovation team has developed tools for automatic detection of doors and anonymization of plates and faces on panoramic images.
Data: MMS point clouds, MMS panoramic images.
Land use planning
GEOFIT works on many topics in land use planning field. The innovation has set up various tools for the processing of terrestrial and aerial data, either images or point clouds, to help the production of land use maps, digital terrain models, automatic classification by theme and automatic vectorization.
Data: multi-resolution aerial images (GSD: 3, 5, 10, 20 cm), terrestrial point clouds (LIDAR MMS, photogrammetry) and aerial LIDAR (5 pts/m² to 100 pts/m²).
Infrastructure monitoring
Our innovation team has developed a tool to detect defects (concrete pathologies) in civil engineering field to simplify the monitoring of buildings and engineering structures. This tool is based on a Deep Learning model for object detection.
Data: drone images.
Underground network
Our innovation team has developed RAPHAL, a subscription-free service that simplifies 3D surveys, and georeferencing. Acquisition is done in the field, using a smartphone or a tablet, and the online platform allows to analyse the acquired data in a collaborative way and in real-time.
Data: smartphone and tablet images.
Our tools and solutions
