• Artikkelin kategoria:Tiedotteet
  • Artikkeli julkaistu:29.11.2024

Tiedote: Nordic webinar on GeoAI

A Nordic webinar on the use of Artificial Intelligence (AI) in relation to geodata – GeoAI – is to be held on December 3rd 2024.

Due to different time zones the webinar starts in Iceland at 1:00 PM (GMT), in Sweden, Denmark, and Norway at 2:00 PM (GMT+1), and in Finland at 3:00 PM (GMT+2).

GI Norden arranges the free webinar with speakers from most of the Nordic countries.

The webinar will show how AI is applied in connection with GIS and geodata in the various Nordic countries. We will see examples from the public sector, from private companies, and from universities.

Among the topics for the webinar that we are working on are Generative AI for Societal Analysis and Urban Planning, the use of multiple input sources for road detection with deep learning neural networks, GeoAI for proactive road maintenance, and a comparison of building extraction using aerial and LiDAR DSMs with deep learning. Besides these applied examples we also intend to discuss why AI is the next step in digitalisation,

Join this webinar to streghten your knowledge on “how to GeoAI”.

Program

02.00 (GMT+1): Welcome to the GI Norden webinar. Miika Kostamo, GeoForum Finland

02.05 – 03.45: Five presentations on various topics within GeoAI from the Nordic countries will be given

02.05: “AI-klivet” – Open data, generative AI, and data visualizationAnders Elias, AI Arena, Sweden
I am an accomplished Open Data Specialist and course leader for a pioneering course on Generative AI for Societal Analysis and Urban Planning. My expertise in digital tools and methodologies extends to driving growth and success within organizations. My extensive experience in business intelligence strategy, data science, AI, and visual analytics equips me to support both public and private sectors in harnessing data for operational enhancement and digital progress. I excel in initiating organizations on their AI journeys, exploring innovative paths in digital transformation, and now, educating professionals in the application of generative AI for comprehensive societal and urban planning strategies.

02.25: Comparison of building extraction using aerial and LiDAR DSMs in urban and rural areas with deep learningEmilia Hattula, Finnish Geospatial Research Institute FGI – National Land Survey, Finland
The presentation will show how the use of light detection and ranging (LiDAR) DSMs is different from the use of aerial DSMs in building extraction using the UNet model. The test included both urban and rural areas in two Finnish cities, Savonlinna and Pudasjärvi. Results with different UNet models trained with building vector labels, DSMs, DEMs and true orthophotos from multiple areas of Finland, were produced with different data combinations to see, which data combination would lead to the most accurate building detection results. Results were evaluated mainly by visual inspection, but in the forest area also quantitative assessment was used. Based on the tests carried out, combining the information from true orthophotos with LiDAR DSMs and 25 cm DEMs provided the most accurate results.

02.45: On the intersection between geospatial and AI. Alexander Nossum, Norkart, Norway
How can we capture more up to date ecological data on a national scale? Is it possible for a geospatial foundation model? The public sector has a tighter economy, so how can we make the sector more efficient in planning and building work? And most importantly, in all efforts – how do we measure the quality of geospatial AI models?
Alexander will guide us through ongoing projects and efforts in the intersection between geospatial and artificial intelligence.

03.05: AI for proactive road maintenance – an ESA projectJohan Bender, Pluto Technologies, Denmark
Since the Romans laid roads across Europe, road engineers have carried out systematic inspections of road conditions. Visual inspections are still used, but this method has its limitations – not least because it is resource-intensive to implement. Together with the European Space Agency (ESA), we are participating in a project to better utilize data from ongoing road surveys with AI. Accurate geographical positioning and registrations opens for automated inventory counting, such as signage, which was previously done manually. During the presentation, you will hear about the results from the municipalities that have participated in the project. You will also hear about how it is possible to build a common, transparent data basis for the municipality to work from in GIS, and about proactive decision-making processes in professional systems. Finally, you will hear about the challenges and opportunities that come with the implementation of artificial intelligence in workflows.

03.25: Enhancing Road Extraction Accuracy with Aerial Lidar Products and Deep LearningJere Raninen, Finnish Geospatial Research Institute FGI – National Land Survey, Finland
The presentation will cover the use of multiple input sources for road detection, including RGB information from aerial true orthophotos, and lidar data products as an additional color channel, including aspect, DTM, hillshade, intensity, roughness, and slope, with deep learning neural networks (RoadVecNet and UNet3+) for road extraction. Lidar data products are not affected by shadows and add 3D information for the neural network, improving the road extraction capabilities, especially in detecting roads under vegetation and shadows. The images and Lidar data are collected by the National Land Survey of Finland (NLS). The dataset contains 23 areas, including both urban and rural areas from different parts of Finland covering a total of 864km2. The accuracy of the predictions with different lidar data products using F1-score, recall, and precision. The results show that integrating lidar data products with RGB imagery enhances road extraction performance.

03.45 Discussion

04.00: Webinar ends

Host and moderator of the webinar
Host of the webinar: Miika Kostamo, GeoForum Finland
Q&A manager of the webinar: Sofi Almqvist, Geoforum Sweden

Registration
All members of one of the Nordic geofora (GeoForum Finland, Geoforum Sverige, GeoForum Norge, LISA Iceland, or Geoforum Danmark) will be granted access for free upon registration.

The webinar takes place in the Captivate/Livehouse (a Danish webinar platform). A link for sign-up and subsequent participation will be sent to registered participants in the morning at the day of the webinar.