How can you extract more information from satellite images than you can with standard band combinations and spectral indices? In this Blended Learning course, you’ll learn to recognize spectral patterns and characteristics and convert them into useful geoinformation. You’ll delve into spectral analysis and feature extraction to analyze differences in vegetation, water, soil, materials, and land cover more accurately.
During this Blended Learning course, you’ll delve into analyzing spectral information and deriving features from satellite images. You’ll learn how different surfaces and materials differ spectrally from one another and how to select and derive relevant features for further analysis. This takes you a step beyond standard image interpretation, band combinations, and spectral indices.
Materials on the Earth’s surface each absorb and reflect electromagnetic radiation in their own unique way. This creates a characteristic spectral pattern for vegetation, water, soil, minerals, and man-made materials. These patterns form the basis for advanced spectral analysis.
You will learn how to interpret spectral curves and reflectance profiles, and how absorption and reflectance characteristics provide information about the properties of the Earth’s surface. You will also investigate which parts of the spectrum best distinguish between different materials and land-cover classes.
You will also learn to account for factors that influence spectral values, such as the atmosphere, moisture, vegetation structure, soil conditions, shade, and acquisition conditions. This will enable you to assess when a spectral difference actually provides meaningful information about the object or material being studied.
A spectral signature describes how a particular material or type of land cover responds to different wavelengths. You’ll learn to derive, compare, and interpret spectral signatures from satellite images.
For example, you’ll investigate why healthy and stressed vegetation can exhibit different spectral patterns, how water differs from moist soil, and why different soil or material types can differ spectrally from one another.
By comparing spectral signatures from known locations or reference data with unknown areas, you can investigate where similar characteristics occur. In this way, spectral signatures form an important link between satellite observations and the physical properties of the Earth’s surface.
Satellite images often contain much more information than is needed for a single analysis. Feature extraction focuses on deriving and selecting characteristics that are relevant to the problem at hand.
You will learn how original spectral bands can be supplemented with derived features, such as band ratios, spectral properties, texture features, and spatial patterns. You will also explore how to reduce redundant information and how to combine different features into a compact dataset that is better suited for further analysis.
In doing so, you’ll be introduced to dimension reduction techniques, such as Principal Component Analysis (PCA). This allows information from multiple correlated bands to be summarized into a smaller number of components that reveal important variations in the satellite data.
Not every object can be reliably identified based solely on spectral values. Two types of land cover may appear similar spectrally, while their shape, structure, or spatial context differs significantly.
That is why you will learn to combine spectral characteristics with spatial information and texture. For example, you will investigate how homogeneous agricultural fields, urban structures, forests, or other landscape elements provide additional characteristics that can help distinguish patterns more clearly.
This results in a richer description of satellite images that considers not only the value of individual pixels but also the spatial structure and coherence within the image.
The features derived during feature extraction often serve as input for subsequent analyses. They can be used, for example, for image classification, change detection, pattern recognition, and machine learning models.
You will therefore not only learn to calculate features but, more importantly, to assess which features actually add information to the problem at hand. In doing so, you will compare different features, investigate their interrelationships, and prevent an analysis from becoming unnecessarily complex due to large numbers of similar variables.
You’ll work with open satellite data and open-source techniques. QGIS serves as a key working environment for combining, analyzing, and visualizing data. Where necessary, additional open-source tools and algorithms can be used for specialized spectral analyses.
During the Blended Learning program, you’ll work with realistic satellite data and real-world problems. You’ll investigate which spectral and spatial characteristics are needed to make differences in the Earth’s surface visible and measurable.
You’ll work on assignments such as:
Upon completion, you will be able to conduct an in-depth analysis of spectral satellite data and derive relevant features from images. You will be able to interpret spectral signatures, develop and select various features, and combine spectral, spatial, and textural information. This will enable you to translate complex satellite data into a usable set of features for further remote sensing analyses.
You’ll receive 1-on-1 guidance. After signing up, our course coordinator will contact you to schedule your first session.
Do you have questions about the course content? Or are you unsure whether the course aligns with your learning goals or preferences? Would you prefer an in-house or private course? We’d be happy to help.
Vegetation has characteristic spectral properties that vary depending on factors such as plant species, growth stage, moisture content, and health. By analyzing spectral signatures and derived features, differences within agricultural fields or natural areas can be identified. This can be used to detect crop variations, vegetation stress, and changes in vegetation development.
Different soil types, minerals, and rocks can have different spectral properties. By analyzing specific parts of the spectrum and characteristic absorption and reflection patterns, spatial variations in soil and subsurface materials can be investigated. Spectral analysis can thus support soil mapping, geological research, and the identification of material differences on the Earth’s surface.
In urban areas, different materials can appear very similar spectrally. By combining spectral information with characteristics such as texture, shape, and spatial patterns, it is possible, for example, to better distinguish between buildings, roads, vegetation, and other urban structures. These features can then be used for classification, monitoring urban development, and analyzing changes in infrastructure.
Spectral, spatial, and texture features can be used as input for classification methods and machine learning models. By first deriving relevant features from satellite images, a model receives information that helps distinguish between different objects, materials, and land cover types. Feature extraction thus constitutes an important step between raw satellite data and advanced applications such as automatic image classification, pattern recognition, and GeoAI.
