Outline of topics covered in remote sensing.
Remote Sensing data
- data collection
- airborne sensor (planes, etc) or satellite (geostationary or orbiting)
- active sensors – emit radiation and record reflection. most common
one is radar, can be directed (SIR-generated DEM)
- passive sensors – rely on sun’s radiation and emission from surface
- sensor types
- pushbroom
- rotating mirror
- spectral sensor type
- panchromatic (SPOT and TM7, band 8)
- multiple, narrower bands
- TM Band 6 is lower resolution
- errors due to
- sensor calibration differences
- scan line shift
- data types
- scale – resolution and scale inverse, can’t have both
- large pixel size, broad coverage (1-10 km grids, weather, global land
characterization, ocean temp/chlorophyll mapping)
- small pixel size, detailed coverage (TM for landuse, spy sats, construction/planning,
30 to <1 m)
- image type
- panchromatic
- multispectral
- hyperspectral
- EM spectrum, radiation and earth
- radiation generated at sun according to temperature
- max radiation at visible wavelengths (measure in micrometers)
- most sensors visible and longer (IR à microwave)
- in atmosphere radiation is
- absorbed and reradiated (e.g., blue sky)
- absorbed (like CO2, H20)
- reflected (like UV)
- refracted
- transmitted
- RS technology for land applications has to work in “atmospheric windows”
which are wavelengths that penetrate atmosphere
- at surface radiation is
- absorbed and reradiated (e.g., in TM band 6)
- absorbed
- reflected
- refracted in water, for example
- the “color” of an image is a combination of what is reflected and
absorbed
- each material on earth has it’s own “spectral signature” which can be
found with a handheld or airborne radiometer (device to measure reflectance-emittance
across a wide variety of wavelengths); can produce characteristic curves
for vegetation, minerals, etc.
- simulated-visual and false-color images
- red, green, and blue color layers of the computer are red, green, and
blue visual light. Whether it “simulates” visual depends on “band” width
versus eye cone sensitivity and “stretch ”routine’s match of brain function.
- any combination (RGB432 for TM is very common) of bands not RGB321 for
TM.
- rectification
- image to image for comparison – not geographically located
- image to map for use in GIS
- ground control or warping points
- mean square error <1 or 0.5 cell size
- Warp “order” degree of distortion
- orthorectification, accounts for “layover” generated by topography,
requires DEM
Processing remote sensing data
- ·LUT (Look-up table)
- describes how to translate a digital number (DN) received at satellite
- single color (red, greyscale)
- multicolor (rainbow)
- ·Band stretching
- contrast enhancement
- sensor ouputs DN 0-255 (or 1-256) (computers like binary series, 256=28)
- stretch data from actual limits to 0-255 (black - white)
- haze corrections
- “dark pixel subtraction”
- caused by refraction/re-emittance from atmosphere
- illumination of “shade”
- mostly affects shorter wavelengths
- band correlation
- greatest variation on most images due to illumination differences +
major reflection differences
- decorrelation stretch attempts to improve differences in band signals
- ·ratios
- remove illumination effect on similar materials
- need to be atm corrected first
- ·principal component analysis
- decorrelation stretch in multiple dimensions
- determine maximum variability
- principal components defined by eigenvalue matrix. shows which band
contributes to the signal in each PC
- PC1 always illumination
- PC2-6 (for TM) contain other types of variability
- increasing “noise” or random signal at higher PC’s
Image Enhancement
- ·edge detection/enhancement
- high-pass filter
- ERMapper’s directional filtering algorithms
- ·change detection
- raw data from single bands at two different epochs
- averages of multiple bands “ “
- ratios and PC’s “ “
- classify first, then compare (ala Chuck Dietzel)
- ·GPS data acquisition
Classification of remotely sensed data
- ·simple discriminants
- ·supervised and unsupervised
- mode/median/majority filtering of results