Munich, Germany / Geneva, Switzerland – January 20, 2009 – GeoDataNetwork, a geographic data management, acquisition, and processing firm, has utilised Definiens Cognition Network Technology to develop a land-use model of the Canton of Geneva. The project involved the analysis of hundreds of aerial and satellite images in combination with a digital height model of the surface to determine land use in Geneva and its surrounding regions. The Section of Official Survey (SEMO) of the Department of Territory of Geneva will use the model to manage its urban planning projects.
Employing Definiens Developer, Definiens Extension for ArcGIS and Definiens eCognition Server, GeoDataNetwork developed a semi-automated application to analyse a series of high resolution aerial photographs and height information of the Canton’s surface to identify areas of commercial, industrial and residential development. The application automatically extracts features and objects indicative of land use, including green space, buildings and transportation structures. The project required the classification of all objects larger than 2 m2 with strict boundary tolerance requirements. Covering an area of approximately 282 km2, the canton comprises 45 communities, of which Geneva is the largest, as well as agricultural areas, small forests and the 38 km2 Lake Geneva. The SEMO will utilise the completed model to detect changes in soil coverage across the canton. The maps will be updated every four years based on the analysis of new aerial photographs.
“Mapping the canton was a challenging project as the aerial images analyzed were collected in August, leading to dark shadows in the urban areas which are difficult to handle,” said Stéphane Couderq, Director of GeoDataNetwork. “Integrating Definiens software into our workflow enabled us to develop a land-cover image analysis solution that is highly accurate and transferable to other image data. We achieved a high degree of automation in the extraction of detailed land cover information, reducing our workload tremendously.”
In heterogeneous environments, such as dense urban zones, shadows, impervious areas and differences in building size, provide challenges to the deployment of image analysis software. Definiens software combines spectral and elevation data to classify objects and describe their semantic relationships. By examining features in context, the software is able to segment complex scenes and classify a far greater range of infrastructure and natural features than traditional pixel-based solutions. Larger buildings can be detected according to their differential elevations to neighboring objects, while form features help separate roads and streams from other areas based upon their irregular or elongated shapes.
“The application developed by GeoDataNetwork truly showcases our technology’s ability to handle complex land-mapping tasks,” said Ralph D. Humberg, Vice President, Earth Sciences at Definiens. “Our object-oriented approach facilitates the accurate and rapid identification of land use and land cover in a range of challenging environments.”