final Flashcards
how does radar work?
- active form of RS
- measured as a function of the radar pulse travel time
- has to do w the physical surface of earth
radar principles
- firing radar beams - thru pulse generator
- need to catch it and process
- looking at point cloud, not an average of what is going on between bounce points
- atmosphere has no noticeable effect
radar components
pulse generator, transmitter, duplexer, receiver, recorder, post-processing
benefits of radar
- all-time/all-weather capability
- info on roughness at a human scale
- deeper penetration of soil
- vegetation may or may not be a concern
limitations of radar
- expensive
- image data is complex and hard to interpret
- little to no info on surface composition
radar bands
- pulses are sent and received in discrete wavelength bands (Ka, C, L)
passive radar collection
- senses the naturally available microwave energy within their field of view
- energy emitted is a function of the dielectric constant and temp
- dielectric constant: the ratio of a material’s ability to store electric potential to the free space’s ability to store electric potential
– greater for things w higher moisture contenta
active radar collection
what is resolution a function of?
- controls the source and data collection
- resolution is a function of antenna length and pulse duration
SLAR (side looking airborne radar)
- good for regional studies
- side-looking geometry affects how the signal interacts with the surface
– causes geometric distortions
ground range
Rg
distance away from the nadir point (perpendicular to the flight direction)
slant range
Rs
distance along the beam path
azimuth
distance along the flight direction
look angle
(Y)
angle from the vertical to the beam
depression angle
theta
complement to the look angle
swath width
illuminated surface of the ground
pulse duration
T
time of the pulse
determines how large the antenna needs to be because the antenna needs to receive pulses sent far away while the ship keeps moving
incidence angle
angle from vertical of the ground and the beam
background of radar images
- backscatter is a function of the statistical variation of the random heights from the surface
- larger wavelengths are not affected by a small statistical surface roughness (also incidence angle)
corner reflector
an object on the surface with a geometry with respect to the incident energy such that all energy is returned to the antenna (“perfect reflector)
flat perfect reflector
bounce off in the same direction
appears dark bc there is no information
rayleigh criteria
relate approximate surface roughness to amount of backscatter
smooth: H <λ/(8sinθ) - shows dark
rough: H > λ/(4.4sinθ) - shows light
radar polarization
- SLAR systems can commonly transmit and receive in different polarization planes
- interaction w surface features can depolarize the beam
- horizontal send and receive is the strongest bc most things have vertical orientation (therefore, they scatter back most of the energy)
radar pulse geometry
- near-range is received first and then the far-range
- all backscatter within any given zone of the swath width parallel to the azimuth direction is received at the same time
- radar data is measured in the slant range direction
- it is re-projected onto the ground range direction (Rg)
range direction (w equation)
ground range distance?
determines the position from a measurement of the travel time from the antenna to the surface (assuming a fixed depression angle)
Rg = Rs * cosθ
Rg = ((cT)/ 2) cosθ
Rs = slant range distance θ = depression angle
slant range (w equation)
- the distance measured along a line between the antenna and the target
- Rs = (c*T)/ 2
range resolution (w equation)
- direction perpendicular to the flight direction
- the ability to distinguish between two objects in the range direction of the radar return
- only can happen if the received pulse from the object closest to the antenna ends before the returned pulse of the far-range object begins
- rr = (c*t)/(2cosθ)
- shorter pulse duration and smaller depression angles result in better range resolution
result of small depression angle
larger radar shadows
more topography shown
result of short pulse duration
less return and more noise
more detail
azimuth resolution
- azimuth resolution = swath width
RAR
real aperture radar
- one pulse sent, one received
- very course resolution
- azimuth resolution gets worse in the far range
how to improve resolution of RAR
- decrease wavelength
- increase antenna length
- decrease slant range distance
SAR
synthetic aperture radar
- uses the principle of the Doppler shift to track the motion of objects in the azimuth direction through successive, overlapping pulses
- objects in the near range are observed for shorter times that those in far range
properties of near range radar
slant range is smaller
θ is larger
range resolution is worse
azimuth resolution is better
properties of far range radar
Rs is larger
θ is smaller
rr is better
ra is worse
foreshortening
-occurs in uneven terrain
- width of ground objects appears to increase toward the far range
- elevated points are displaced toward the antenna
- a return from the bottom of the feature prior to the top
- illuminated slope lengths appear to decrease
- hard to correct
lay-over
- occurs in uneven terrain
- return from the top of the feature prior to the bottom
- the radar wave front is steeper than the slope angles
- hard to correct
InSAR
Interferometric SAR
- combine two or more SAR images of the same area
- measure the phase change in the returned signal from the surface
- studies earthquakes, volcanoes, glacier flow, etc
null kernel (averaging)
- used for noise/line removal
- kernel is passed over the image center pixel values - which exceed the average threshold are replaced
low pass filter (box car filter)
- kernel averages DN values, which is substituted for the middle pixel value
- removes “high frequency” features
- blurs
high pass filter (non-directional)
- all feature edges are enhanced (increases contrast)
- kernel values sum to 0
high pass filter (directional)
- enhance directional features
- choice of kernel determines which features
- removes all low frequency features and accentuates linear features in the direction of the kernel
principle component analysis
- highly correlated data in two image bands will lie along a line/trend
- if this plotting is carried out for each band, the results will vary from the most correlated in
PC band 1 (topography/temp)
to less correlated (minerology) to the lease correlated in PC band n (noise) - benefits: reduction in data volume, unit discrimination, removal of noise
decorrelation stretch
- follows a principle component analysis to enhance variation
- band axes are rotated into the plane of the eigenvector
- stretch is performed perpendicular to the eigenvector
-stretched data are rotated back - goal is to cover a wider range of DN values and emphasize differences in bands
digital elevation models
stereo pairs: successive overlapping air photos
- works bc each photo images each point on the ground from a slightly different angle, the offsets can reproduce the vertical dimension
- can be done in one pass if the satellite has two antennas
types of DEM
shaded relief maps
- diff shades of gray are assigned to slopes
color density slices
- color-coded elevations
how LIDAR works
light detection and ranging
- transmitted light interacts with and is changed by the target (reflected and/or scattered back)
- timing is measured
- determines the roughness and scattering properties
3 basic types of LIDAR
- range finders
- DIAL (differential absorbers LIDAR)
- Doppler LIDAR
range finder LIDAR
- used to find the distance from the distance to a target
- sending out beams to specific points and registering that info
DIAL
differential absorption LIDAR
- measures chemical concentrations in the atmosphere
- picks one line of longitude and measures it the whole way thru
- good for larger molecules
doppler LIDAR
used to measure the velocity of a target
radiant energy in TIR
- energy emitted from a surface rather than reflected from it
what is Wein’s law and what’s the formula
- linear approximation of the wavelength of the peak energy and the temp of the object
- describes the wavelength shift of the peak emitted energy from a body with changing temp
λ = 2897/T (T is in Kelvin)
what is the planck equation
describes the radiant energy at all wavelengths and temperatures
TIR Emissivity
- emissivity varies with wavelength for most materials
- used for compositional identification
- derived by extracting it from the radiant energy (separated from the temp component)
blackbody
material with no emissivity variations
thermophysical properties
- thermal inertia (how long it takes to heat up and remain hot before cooling)
- thermal conductivity (rate of heat flow through a material)
- density
- heat capacity (heat storage of a material)
what composes radiation
temp and emissivity
reference channel
- assume a max emissivity and assign to wavelengths
- derive the temp at the wavelength
- use the temp to derive remaining emissivity values
apparent thermal inertia equation
ATI = (1-A) / (Tday - Tnight)
A = albedo image - broadband surface reflectance
Tday = max daytime surface temp
Tnight = min night time surface temp
emissivity normalization
-calculate 5 reference channel temps
- highest temp is equated to band w highest emissivity
- then reference channel approach
how to train a deep learning model
- prepare training dataset
- train the model
- run the model
- review/validate results
