Classification of clutter data and object height for Icelandic telecommunications company

We support the Islandic telecommunications company Nova through the provision of up-to-date clutter data, detailed object elevation information, and a digital terrain model (DTM) for the entire Icelandic territory (103,000 km²)For the Islandic telecommunications company Nova up-to-date clutter data, detailed object elevation information, and a digital terrain model (DTM) was provided for the entire Icelandic territory (103,000 km²).

In the context of radio network planning, clutter data refers to information about physical obstructions or environmental features that can affect wireless signal propagation. This data includes details about buildings, trees, terrain, and other objects that may obstruct or reflect radio waves, influencing signal strength and coverage. Clutter data is essential for accurate radio network planning and optimization, as it helps predict signal behavior and identify areas with potential coverage issues. Efficient management and analysis of clutter data are crucial for designing robust and reliable wireless networks. The clutter data classification consists of a 16 class nomenclature, natural classes like water, grassland, forests, and several density classes of built up area.

Since a higher level of detail is required in densely populated areas than in rural areas, the total area is divided into three zones: the city area of Reykjavík, all other urban centres, and the remainder of Iceland. For each zone, the products are offered in a different but adequate resolution in order to provide the best cost-benefit ratio.

The zone of Reykjavíkwas meticulously analysed using very high-resolution datasets, with clutter data sourced from a newly captured orthophoto in July 2023. Digitized data samples were utilized as training datasets for a classification based on an U-Net algorithm –  a convolutional neural network architecture commonly applied in remote sensing for precise image segmentation tasks. This process facilitated an intricate land cover/land use classification at a very high spatial resolution of 1m. Additionally, height models, specifically the Digital Terrain Model and Digital Surface Model, were generated from stereo-orthophotos, delivering current and detailed elevation data for both buildings and vegetation in the area.

A thorough visual inspection was conducted across the entire country to ensure complete building coverage. Building heights were determined from the corresponding nDSM, while the building classification process involved density analysis and supplementary layers for accurate identification.