Advanced Geoinformatics

Image Analysis

In this course, you will be introduced to advanced image analysis methods enabling to enrich your geoinformation problem solving abilities. Non-linear filters will be introduced for reduction of noise while preserving the boundaries. In addition, interest operators will be introduced to detect stable structures in images that are invariant to scale and rotation transformation. Various methods for dealing with objects in images will be studied. Fuzzy and sub-pixel classification will be introduced to deal with uncertainty and to increase the information content extracted from the imagery. For multisource classification decision trees will be introduced. To automatically detect corresponding image positions, the image matching techniques will be introduced. In particular, area-based matching and feature-based matching will be investigated in this course.

Laser Scanning

Airborne, terrestrial and mobile laser scanning are modern technologies to acquire and monitor the geometry of the Earth's surface and objects above the surface like buildings, trees and road infrastructure. This course provides an overview on the state of the art of these techniques, potential applications, like digital terrain modelling and 3D city modelling, as well as methods to extract geo-information from the recorded point clouds.

Integrated Geospatial Workflows

Thanks to the digital, mobile and IT revolutions, massive amounts of data are nowadays collected at unprecedented spatial, temporal, and thematic scales by both physical and human sensors. For most applications, data availability is less of an issue. What remains an issue is how to convert this data into usable and actionable geo-information that supports decision-making at various scales and that can be further processed to generate knowledge. As a consequence, scientific workflows and scientific workflow management systems become more important for knowledge sharing and ensuring reproducibility.

Selected elements of an integrated workflow are introduced. Methods and techniques that can process massive amounts of spatio-temporal data in (quasi-)real time by using cloud computing technologies will be discussed. This course combines and extents knowledge on semantics, linked data, machine learning and distributed databases, and interactive dashboards presented online.

Spatio-temporal modelling and Analytics

Processes relevant to system Earth, whether natural or man-affected, commonly display variations in space and over time, yet our understanding of their behaviour remains limited. The increase in available monitoring data provides handles for detailed study of these processes. Unravelling the way these processes function and having a mechanism to test hypothesis as well as the possible impacts of interventions is key to contribute to a more sustainable development.  At course end, you will have learnt to make use of the available data in process studies, by a variety of computational techniques.

Upon completion of this course, you will be able to:

• Develop an image processing chain using non-linear filters and mathematical morphology operations for automatic information extraction from images in context of a given problem.
• Choose and apply a segmentation method to a given image and describe the uncertainty of the obtained result.
• Make informed decisions on the best classification method for a given set of images and a specific problem.
• Apply orthorectification to derive orthophoto.
• Make informed decisions on appropriate image matching method for a given type of data and problem.
• Evaluate attribute and scale uncertainty and relate it to the quality of derived orthophotos, accuracy of resulting image classification and matching.
• Explain how point clouds are generated from GNSS, IMU, and range finder measurements and relative sensor registration.
• Assess the applicability of laser scanning for various tasks, like surface reconstruction and 3D modelling.
• Design survey plans to acquire point clouds taking into account the accuracy and point density requirements.
• Evaluate the quality of laser scanning datasets.
• Determine and apply optimal point cloud processing methods to extract surface descriptions for geometric modelling and point cloud classification.
• Interpret and analyse point cloud processing results.
  

Upon completion of this course, you will be able to:

• Analyse the quality of structured and semi-structured data sources and apply coding solutions for the storage, querying and curation of this data, appropriate for specific application contexts.
• Apply semantic information integration through knowledge formalisation, semantic enrichment, exploratory querying and data mining.
• Construct interoperable and reproducible geospatial workflows based on process modelling methods and workflow languages.
• Make informed decisions on the infrastructural system design for enabling meaningful data integration on the web.
• Discuss the main spatio-temporal modelling paradigms.
• Design a conceptual model for a spatiotemporal ABM using UML and the ODD protocol.
• Implement a basic ABM model, and calibrate this model using behavioural space.
• Explain to peers the main advantages and limitations of using geo-computational methods.
• Choose and integrate appropriate geocomputational methods to study a simple spatiotemporal problem.
• Organize and conduct the modelling and analysis phases required by a simple spatiotemporal project.
• Apply cloud computing approaches to support and/or realize the main modelling and analysis phases.
• Evaluate the innovation and societal relevance and impact of the project.