5 ECTS credits
145 h study time
Offer 1 with catalog number 4023572DNR for all students in the 2nd semester at a (D) Master - preliminary level.
The course focuses on technical skills in processing of satellite images and current trends in remote sensing. It is not a normal beginner’s course on Remote Sensing. It is designed for students without Geoinformatics/GIS or Remote Sensing background who wants to learn the subject, or students with related background but does not have the necessary processing skills (concept and practical) to deal with the data. The objective is to introduce the increasingly important spaceborne technologies with satellite images, and prepare the students to tackle pressing environmental problems with advanced methodologies: machine learning algorithms, hyperspectral data, texture analysis, image/sensor fusion, etc.
Theoretical parts:
Solar radiation, electromagnetic wave, atmospheric transmission and spectroscopy imaging.
Basic elements of satellite images. Spectral, spatial and radiometric resolutions.
Requirements for pre-processing and processing of remote sensing images
Earth Observation satellite missions and sensors
Spaceborne Earth Observation missions (Landsat, SPOT, ASTER, Sentinel, etc) and sensor characteristics
Showcase of application: urbanization, environmental monitoring, natural disaster, agriculture, forestry.
Radiometric correction, geometric correction, atmospheric correction
Classification of satellite images. What is land use/land cover classification?
In-depth study of procedure/protocol for satellite image classification.
[Classification scheme (Levels) -> Training pixels -> Separability test->feature selection->classification approaches->Accuracy assessment.]
Supervised and unsupervised methods and accuracy assessment.
Real case study demonstration
Classical change detection techniques will be introduced. Image Differencing, Image Ratioing, Post-classification comparison, Change Vector, Temporal spectral clustering
Classical texture measures: Grey Level Co-occurrence Matrix, Textural Spectrum, Semivariogram
Recent development and applications.
Hot research topics related to such data and the implications to future development in remote sensing.
To verify or quantitatively assess a classification map generated from remote sensing images, in-situ knowledge and reference data are needed. Groundtruth generation: reference data, HD aerial photos, manual interpretation. Field campaign/measurements.
Use of Random Forest/Deep Learning based methods for land cover/land use mapping and other environmental applications.
Ecotope Mapping in Natura2000 landscapes. Ecosystem Service. Climate Change Mitigation. Marine Plastic.
Practical parts:
Practical I
Access to open source satellite images/data.
Understanding sensor configurations. Processing levels.
Understanding open-source processing tools (e.g. QGIS)
Practical II
Preprocessing of a satellite image (geo-referencing, image co-registration, pan-sharpening, spectral enhancement).
Practical III
Land cover classification of a satellite image. Accuracy assessment of a classification map.
Practical IV
Change Detection
The use of Image Differencing and Post-classification comparison change detection
Practical V
Application of machine learning algorithms (e.g Random Forest, Deep Learning model) for information extraction and analysis.
..
under construction
The final grade is composed based on the following categories:
Other Exam determines 100% of the final mark.
Within the Other Exam category, the following assignments need to be completed:
..
This offer is part of the following study plans:
Master in Applied Sciences and Engineering: Applied Computer Science: Standaard traject