remote sensing midterm 2

image classification

assigning individual pixels to a target catetory or themetic information

3 things image scheme should be

exclusive

exhaustive

heirarchical

spectral seperability

the idea that pixels can be seperated into different categories based on spectral wavelengths, but additional bands may be necessayr

supervised

some prior knowledge by mapper

training data created

training data used to classify all

unsupervised

unknown

computer randomly generates clusters

clusters are relabeled into categories

training site statistcis

what should be analyzed to determine if the different categories are discriminable

feature space plots

depic the distribution of all scenes

minimum distance classifier

using nearest mean to determine

ecluidan distance

parellelpiped

uses mean and SD to create like boxes pixels are put in closest box

nearest neighbor

pixels are put with nearest training data

maximum likelihood

based on probability / probability calculated from density functin

unsupervised classificaiton

isodata

mixed methods

perform unsupervised on known regions

what needs to be considered when analyzing changes in a region over time

spatial study domain - temporal study domain - resolution - comparable data

image algebra

taking pixel by pixel ratios to identify changes between two calibrated images

common with NDVI or vegetation indicies

thershold determination

analyzing when a distribution appears to be not following a normal distribution curve anymore

set that region for change as opposed to no change

anomoly plots

image of one time period compares to average as opposed to change over time

multitemporal color comp

mixing dates of the same spectral band to see differences

normalized difference wetness index

multidate layer stacking

stacking multiple dates to view change

spectral change vector analysis

for every pixel, give it vector and magnitue of change, map what direction pixels change and how intense

time series analysis

looking for specific change points in a curve of data

satellite based thermal IR radiation

aim of thermal IR radition is often to derive surface temperature

when temp is inferred from measurement of emitted radiation it is called

radient temperature

correction factor

emissivity

emissivity is

ratio emitted relative to what is expected with a blackbody

an objects emission follows a

characteristic curve

what happens to any object above 0

it emits electromagnetic radiation

what is kinetic temperature

the objects actual physical temperature

what is an ideal blackbody

an ideal emitter that emits radaition according to black body radiation laws and has emissitivity of 1

kirchoff law

spectral emissivity is generally equal to spectral absorptance

peak wavelength of earth

10 micrometers in the thermal infared

when are radiant temp and ideal temp equal

in an ideal emitter

what happens if emissivity is underestimated

temp will generally be underestimated

around what microns is the ideal estimater generally good and not as close to 1

5 good 11 far from 1

for infared thermal we can assume earths surface is

opaque

LiDar stands for

light detection and radiation

why use lasers for lidar

laser is concentrated in one direction

energy is concentrated around one wavelength

what does active remote sensing require

transmitter and reciever

what does lidar collect

backscatter profiles of altitude profiles

height data

t = time it takes for pulse to travel from sensor to site and back to sensor c = speed of light target elevation = instrument altitude - range

divide by 2 to measure time round trip

nadir and scanning

profiling lidar and imagind lidar

nadir

measuring distances directly beneath the aircraft

targets only areas along the flight path

scanning lidar

laser sweeps across the landscape and different angles

wider more complete area

footpring size

area illuminated by an individual laser pulse on the ground

depends on divergence and height above ground

increasing height generally increases footpring size

footpring spacing

how far apart are the individual measurments

discrete return lidar

managable data volume

low verticle resolution

straight forward to process

first return, one or more intermediate, last

full waveform lidar

informs more about how vegetation is distributed vertically

preserves continuous returns

requires more complex processing

point clouds

collection of points representing the three dimensional positions of detected surfaces

digital surface model

natural and built aspects of the surface

represents first reflected surface

digital terrain model

more useful in hydrology or soils

focused on last returns

differential absorption lidar

materials absorption depends on wavelength

need multiple bands

GEDI

global ecosystem dynaomics investigation

high resolution lidar instrument designed to measure earths forests and topogrophy

ICEsat and GLAS

radar

have a broad atmospheric window,

can be used to target different ranges

shorter radar pulses mean

better range resolution

slant range

acttual distance sensor ro object

ground range

sensor to point between distance to object

foreshortening

compressing, making top come before bottom

layover

top returns before bottom