Durchschnittlicher Fortschritt
Empfohlene Karteikarten-Sets
Karteikarten im Karteikarten-Set
Noch nicht gestartet (8)
What is the spectral signature?
the unique way of absorbing reflecting and transmitting in different wavelength regions of a surface material
How is the spectral detail measured by a hyperspectral sensor different from a multispectral sensor?
hyperspectral sensors capture hundreds of narrow bands -> resolve subtle absorption features
multispectral sensors capture a few broas bands -> resolve broader spectral features
Why do we work with multispectral data from Landsat and Sentinel nevertheless?
their high temporal resolution
To map land cover we can assign categorical labels to each pixel describing their spectral signature as a land cover type.
What is the teoretical feasability of seperating land cover classes governed by?
- the spectral and radiometric detail provided by the sensor (more detail -> more differentiation)
- the spectral (dis)similarities between classes
When is a land cover class catalogue considered complete?
when there are no major surface types omitted that are not included in the categories
What is the difference between thematic and spectral classes?
Thematic land cover and land use classes consist of multiple spectral classes. Spectral classes are clusters of several surfaces with similar spectral properties. It is possible that several thematic classes share similar spectral classes -> hence are not (easily) seperable
What is training data?
the data we use as reference to train our model:
chosen points for which we can determine land cover -> later trains the model to predict land cover on its own
What needs to be consideres when collecting training data?
- A precise and robust definition of your target classes
- high quality reference information if possible
- Stratify your samples - spatially, spectrally (and temporally)