Land Mapping Flashcards
What is the Normalized Difference Vegetation Index (NDVI)? On which special feature of green vegetation is it based? With satellite images, why can’t we use the spectral signature of green vegetation directly?
- Spectral transformation of 2 or more bands designed to enhance the contribution of vegetation properties and allow reliable spatial and temporal intercomparisons of terrestrial photosynthetic activity and canopy structure variations.
- NDVI = (NIR – R)/(NIR + R) (Normalized difference vegetation index)
- It is based on the Reflactance of R and NIR. The higher the difference, the better the recognition of the vegetation
- We can’t use the signature directly because we only get specific bands of electromagnetic radiation. In this case R and NIR
calculate NDVI and determine the band color and if they are vegetated or water?
List and explain at least two properties of satellite systems
Spatial properties:
Orbit Geometry/ Coverage area/ Revisit time/ Spatial resolution
Orbit geometry:
* A satellite in orbit about a planet moves in an elliptical path with the planet at one loci of the ellipse. Important elements of the orbit include its altitude (more than 400 km above the earths surface), period (time to complete orbit is related to its altitude), inclination (refers to to the angle at which it crosses the equator) and equatorial crossing time.
* sun-synchronous orbit:
- These orbits are designed so that the satellite passes over any given point on the Earth’s surface at the same local solar time during each orbit.
- typically polar orbits, which means they pass over the Earth’s poles as they orbit.
- used for satellites that need consistent lighting conditions for imaging purposes, like capturing images of the Earth’s surface for remote sensing.
- Because they follow the Earth’s rotation and orbit the Earth at an angle, they can capture images with consistent lighting conditions, making it easier to compare images taken at different times
- geostationary orbits:
- orbit the Earth at the same speed that the Earth rotates, which allows them to remain fixed over the same point on the Earth’s surface.
-They orbit at a very high altitude, about 35,786 kilometers above the equator. - ideal for applications such as weather monitoring and telecommunications because they can provide continuous coverage of a specific area on the Earth’s surface.
- Since they stay fixed relative to the Earth’s rotation, they’re great for tasks that require constant observation of a specific region, like tracking weather patterns.
Coverage area
spatial resolution:
* The nadir (direction pointing directly below a particular location) is determined by the orbital altitude and the instantaneous field of view.
Revisit time
_________________________________
Spectral p:
*Width of bands/ Numbers of bands/ position of bands
* The extreme cases are panchromatic
(black & white) sensors with a single broad spectral band and hyperspectral sensor with a hundred or narrower, more continuous spectral bands
___________________________________
Radiometric p: Radiometric resolution/ Gain settings/ Signal to noise ratio
*The radiometric properties of a remote
sensing system include the radiometric resolution, typical expressed as the number of data bits used to record observed radiance. They also determine the gain of settings that determines the
range of variation in brightness over which the system will be sensitive, and the signal to noise ratio of the sensor.
*It should be noted that increasing the spatial and spectral resolution, both
result in a decrease in the energy available to be sensed. A high spatial resolution means that each detector is receiving energy from smaller area, while a high spectral resolution means that
each detector is receiving energy in a narrower range of wavelengths. Thus, there are trades off among spatial, spectral and radiometric properties.
Discuss the characteristics of typical orbits of sun-synchronous and geostationary remote sensing satellites.
sun-synchronous:
It’s an orbit that results from a combination of orbital period and inclination in such that satellite keeps pace with sun´s westward progress as the earth rotates. Thus, the satellites crosses the equator at precisely at the same local time as sun time.
- Lansat sensors are sun synchronous, cross equator at an angle of about 9° from normal, 2760 km
Geostationary:
An equatorial orbit at the altitude of approximately 36000 km that produces an orbital period of exactly 24 hrs. It completes an orbit around the earth in the same time needed for the earth to rotate.
Explain the supervised and unsupervised classification in remote sensing. (supervised)
supervised classification:
Identify —> classify
1. Training stage (ground truth):
the analyst identifies representative training areas and develops a numerical description of the spectral attributes of each landcover type of interest in the scene.
- classification stage:
- each pixel in the image data set is categorized into the land cover class it most closely resembles. If the pixel is insufficiently similar to the training dataset, it is usually labeled as “unknown”.
- The category label assigned to each pixel in this process is then recorded in the corresponding cell of an interpreted dataset (an “output image”). –> the multidimensional image matrix is used to develop a corresponding matrix of interpreted land cover category types. - output stage: After the entire dataset has been categorized, the results are presented in the output stage (3). Three typical forms of output products are: thematic maps, tables of full scene or subscene area
statistics for the various land cover classes, and digital data files amenable to include in
a GIS (GIS inputs).
*Among the methods of land cover classification, some examples are:
Minimum distance to the means classifier:
The mean or average, spectral value in each
band for each category is determined. These values comprise the mean vector for each
category. A pixel of unknown may be classified by computing the distance between the
value of the value of the unknown pixel and each of the category means. After
computing the distance, the unknown pixel is assigned to the “closest” class. If the pixel
is further than an analyst defined distance from any category mean, it would be
classified as unknown. The minimum distance to means strategy is mathematically
simple and computationally efficient, but it has certain limitations. It is insensitive to
different degrees of variance in the spectral response data, therefore, is not widely used
in applications where spectral classes are close to one another in the measurement
space and have high variance.
Parallelepiped classifier: We introduce sensitivity to category variance by considering
the range of values in each category training set. This range may be defined by the
highest and lowest digital number in each band and appears as a rectangular area in the
two-channel scatter diagram. An unknown pixel is classified according to the category
range, or decision region, in which it lies or as “unknown” if it lies outside all the regions.
The multidimensional analogs of these rectangular areas are called parallelepipeds. It is
also very fast and efficient computationally. Difficulties are encountered when category
ranges overlap. Unknown pixel that occur in the overlap areas will be classified as “not
sure” or be arbitrarily placed in one or both of the overlapping classes. Overlap is caused
because category distributions exhibiting correlation or high covariance are poorly
described by the rectangular decision regions (Covariance is the tendency of spectral
values to vary similarly in two bands, resulting in elongated, slanted clouds of
observations in the scatter diagram). In the presence of covariance, the rectangular
decision regions fit the category training data very poorly. Spectral response patterns
are frequently highly correlated and high covariance is often the rule. To alleviate, the
parallelepiped classifier can be modified to single rectangles in a series of rectangles
with stepped borders. These borders describe the boundaries of the elongated
distributions more specifically.
Gaussian Maximum Likelihood classifier: It evaluates both, the variance and covariance
category spectral response patterns when classifying. The assumption is that the
distribution of the cloud of points forming the category training data is Gaussian
(normally distributed). If this assumption is reasonable, the distribution of a category
response pattern can be completely described by the mean vector and the covariance
matrix. Given these parameters, we may compute the statistical probability of a given
pixel value being a member of a particular land cover class. The resulting bell-shaped
surfaces are called probability density functions (pdf), and there is one for each spectral
category. The probability density functions are used to classify an unidentified pixel by
computing the probability of the pixel value belonging to each category (the probability
of the pixel value occurring in a particular distribution of the class). After evaluating the
probability in each category, the pixel would be assigned to the most likely class (highest
probability value) or be labeled as “unknown” if the probability is below the threshold
set by the analyst. The maximum likelihood classifier delineates “equiprobability curves
contours” in the scatter diagram, which shape expresses the sensitivity of the likelihood
classifier to covariance. The principal drawback is the large number of computations
required (slower).
Explain the supervised and unsupervised classification in remote sensing. (unsupervised and hybrid)
classify –> identify
*It does not utilize training data.
* involves algorithms that examine the unknown pixels in an image and aggregate them into a number of classes based on the natural grouping or clusters present in the image values.
*The basic premise is that values within a given
cover type should be comparatively well separated.
*The classes that result are spectral
classes.
*Because they are based solely on the natural grouping, the identity of the
spectral classes will not be initially known.
*The analyst must compare the classified data
with some form of reference data to determine the identity and information value of
the spectral classes. The classifier scene may be so numerous that it would be difficult
to train on all of them, and in the unsupervised approach they are found automatically.
hybrid classification:
Unsupervised training areas are image subareas chosen
intentionally to be quite different from supervised training areas.
classification Matrix
- Training set pixels that are classified into the proper land cover categories are allocated along the major diagonal of the error matrix.
*All non-diagonal elements of the matrix represent errors of omissions or commissions. *Omission errors correspond to non-diagonal column elements (16
pixels were omitted from “S” classification).
*Commission errors are the non-diagonal row elements (38 “U” pixels plus 79 “H” pixels were improperly included in the “C” category).
*Overall accuracy is computed by dividing the total number of correctly classified pixels (sum of the elements of the diagonal) by the total number of reference pixels.
*Producer’s errors (Error type II) results from dividing the number of correctly classified pixels in each category (diagonal) by the number of training pixels (column total).
*User’s accuracy (error type I) are computed by dividing the number of correctly classification pixels (diagonal) by the total number of pixels classified in that category (row total). (These procedures just indicate how well the statistics
extracted from those areas can be used to categorize the same areas and give no information about how well the training areas were taken (based on how homogeneous training classes are separable, good strategy))
How to calculate the change in vegetation from false color image and Why are false colors used to display land uses?
and water.
True colors:
Red: Landsat Band 3 (red)
green: Landsat Band 2 (green)
blue: Landsat Band 1 (blue)
False colors:
Red: Landsat Band 4 (Near Infrared)
green: Landsat Band 3 (red)
blue: Landsat Band 2 (green)
False color composites allow us to visualize wavelengths that the human eye can not see (i.e. near-infrared). Using bands such as near infra-red increases the spectral separation and often increases the interpretability of the data. Using false colors for land cover allows us to differentiate the vegetation components.
e.g. Forest has strong IR bands (???) and displays red, making it easier to distinguish from representation of soil
What are the types of projections of earth:
types of projection:
- Equal area → preserves the area not the shape
-Conformal → preserves the shape not the area e.g. Lambert / Mercator
