L3 Spectral Reflectance Flashcards
What is meant by spectral reflectance?
For a given substance/surface, the amount of solar radiation absorbed, transmitted or reflected varies with wavelength. The balance between all of these determines the surface’s colour
What is a spectrometer?
This is a device that plots a graph showing the amount of radiation that a substance reflects in the different wavelengths
What is the red edge?
This is the signature of vegetation on a spectrometer in which it does not reflect very much red visible light but does reflect a lot of infra-red radiation. The transition between them appears as a sharp edge on the spectrometer.
Why does vegetation not reflect much red visible light?
Because it needs to absorb it for photosynthesis
Why does vegetation reflect a lot of infra-red radiation?
Because photons in the longer infra-red wavelengths do not carry enough energy for photosynthesis
What is the calculation for vegetation index and what is it a measure of?
Infra-red radiation reflection - red visible light reflection
What is a slight problem for the vegetation index?
aspect affects the illumination of surfaces
What is the Normalised Differences Vegetation Index (NDVI) a measure of?
the amount of photosynthetic activity present in a plant
What is the calculation for NDVI?
(NIR - RED VL) / (NIR + RED VL)
What is the Normalised Difference Snow Index (NDSI) used for?
Help distinguish clouds from ice in polar areas
How do you calculate NDSI?
(SWIR - RED VL) / (SWIR + RED VL)
What function do feature spaces provide towards spectral reflectance?
they are essentially the graphs in which the characteristics of different surfaces at different wavelengths are presented. You can combine these characteristics which are presented visually to identify what different surfaces are
What is the common minimum number of bands that different satellites have and what does this provide towards understanding the spectral reflectance of different surfaces?
They usually have 7 or more which means we can detect how surfaces respond to 7 different wavelengths which is a lot of information we can use to detect different substances.
What is the other way we can measure spectral reflectance other than % of reflectance?
Euclidean distance in a dimensional features space
What happens when you use Euclidean distance method to identify different surfaces on a computer?
Image classificiation
What are the two forms of image classification?
Supervised and unsupervised
What are two benefits of classifying images?
it makes images easier to determine and interpret qualitatively and may reveal hidden information
What is unsupervised classification?
Using spectral information to group pixels with similar spectral properties in to ‘clusters’. The computer does the work after you tell it how many clusters you want to identify
What are 3 pros of using unsupervised classification?
No prior knowledge required
Low opportunity for human error
Unique classes recognised as distinct units
What are 3 cons of using unsupervised classification?
spectrally homogenous classes may not correspond to informational categories of interest
limited control over classes
spectral properties of specific informational classes will change with time infringing upon ability for temporal comparison
What is the process of unsupervised classification?
1) user specifies number of clusters
2) computer selects centroid positions for clusters based on euclidean distance for each pixel to centroid position
3) computer calculates new centroid position based on mean value of each cluster pixel then adjusts the euclidean distance calculation
4) process repeats until adjusting of pixels position to centroid position no longer changes
What is supervised classification?
When images are classified based on information the user provides before the process begins
How does the box classifier supervised classification work?
1) identify number of classes based on prior knowledge
2) create training sites for each class and define the spectral signature requirements for each
3) computer assigns pixels to each class
4) Each pixel is assigned to the different classes
What happens if the pixel lies in the overlap between 2 or more boxes during box classifier supervised classification?
It uses minimum distance from the mean of each cluster to decide which class it should join
What happens if the class box seems unrepresentative of te pixels distribution within it during box classifier supervised classification?
Adjust them using a parallel piped or box classifier
What are 2 pros to the box classify supervised classification method?
Easy to program
fast in operation
What are 4 cons of the box classify supervised classification method?
1) min and max values may not be truly representative
2) no other information used
3) the shape is not ideal for the nature of the data
4) potentially does not assign all pixels
What is the process of minimum distance supervised classification?
1) finds cluster centroid positions
2) finds the closest cluster for a pixel based on its central position, even if the boundary for that cluster is much closer
What are the pros of minimum distance supervised classification?
1) fast in operation
2) assigns all pixels
What is the con of minimum distance supervised classification?
Ignores class variability
What is the process of maximum likelihood supervised classification?
location, shape and size of clusters determined from statistical properties of training data.
What is the maximum likelihood supervised classification regarded as?
Probably the best method due to its complexity and ensuing accuracy
What is the maximum likelihood supervised classification method based upon?
Probability - how likely pixels are to be attributed to different cluster
What is the pro of the maximum likelihood supervised classificaiton method?
usually the most accuratte
What are 3 disadvantages of the maximum likelihood supervised classification method?
1) high computational cost
2) input data must be normally distributed
3) over classifies classes with large dispersion