Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
137 Cards in this Set
- Front
- Back
|
Draw a simple depiction of the remote sensing system. |
Sun->Tree->Sensor->House |
|
|
Describe the steps in a remote sensing system. |
1) Energy source 2) Transmission through atmosphere 3) Interaction with target 4) Retransmission through atmosphere 5) Recording of energy by sensor 6) Data download 7) Interpretation and analysis 8) Application |
|
|
Describe the portions of EM spectrum used in remote sensing. (Different types and wavelengths) |
Ultraviolet <0.4 um Visible: Blue 0.4-0.5 um Green 0.5-0.6 um Red 0.6-0.7 um Infrared: Near-IR: 0.7-1.3 um Mid-IR: 1.3-3 um Thermal-IR: >3 um
|
|
|
Describe the relationship between different wavelengths and energy. |
Increase in wavelength = decrease in energy |
|
|
Scattering |
Unpredictable spread of energy by particles in atmosphere. Amount of atmosphere->path length. Interaction depends on wavelength and atmospheric particles. Ex: clouds, energy never touches the ground. Degrades image quality |
|
|
Absorption |
Loss of energy to e.g. particles in atmosphere. Caused by water vapor, CO2, O3. Varies with wavelength. Atmosphere doesn't transmit all wavelengths. |
|
|
Reflectance |
Nature of reflectance varies with surface roughness. Smooth surface-specular reflector. Rough surface-diffuse reflector. To quantify reflectance, it's portion of incident energy reflected. |
|
|
Spectral Reflectance Curves |
In choosing a wavelength to operate at, you want to separate the objects to make them as unique from each other as possible. Choose wavelength with max difference between the two |
|
|
Airborne vs. Satellite Platform |
Improved vantage point Unbiased nature of aerial imagery Time related considerations (permanent record of current conditions, synoptic view) Stereoscopy (possible 3-D images, considers height/elevation changes) Increased spatial resolution Airplane sensors get more energy than satellite and quicker data processing. |
|
|
Photographic Film Types: Compare panchromatic to B/W infrared. |
Panchromatic operates at a shorter range than B/W infrared. Panchromatic (0.3-0.7) B/W infrared (0.3-0.9) |
|
|
What is the focal length? |
Distance from where light comes in and where the sensor is located inside camera |
|
|
Describe scale. |
Unitless. Is relationship between photo and ground distance. Can be measured by comparing photo distance/ground distance. Scale over variable terrain is focal length/height above avg terrain. |
|
|
Relief Displacement of photos |
Photo scale varies with objects at different elevations that are displaced relative to one another. Can be calculated if known scale and orientation. Layover effect more severe at near range. |
|
|
Orthophotographs |
Remove effects of scale variation, relief displacement of terrain, tilt. Resulting image has geometric properties of map, removes stereo possibilities |
|
|
Stereoscopic Parallax |
Displacement of object position with respect to a reference caused by shift in observation, 2 separate viewpoints. Only care about axis plane is flying on. Flight line not coincident with fiducial lines. Can solve using equal triangles. Use when 2 photos don't match up. It's important b/c it can be used for digital image matching and correcting distortion. |
|
|
What are the four types of resolutions? |
1) Spatial 2) Spectral 3) Radiometric 4) Temporal |
|
|
Describe spatial resolution. |
Relates to pixel size. Pixel size is best case scenario. How close 2 objects can be so that you can still distinguish them. Depends on atmospheric conditions. |
|
|
Describe spectral resolution. |
Relates to wavelengths. How many bands you can put in wavelengths and what width for each band. Relates to wavelength intervals recorded by sensor. |
|
|
Describe radiometric resolution. |
Relates to number of grey-scale values, bits. Replaces color with energy. Shown graphically using characteristic curve. Can detect different types of energy from different colored objects. 2^# of switches 8 bits = 1 byte |
|
|
Describe temporal resolution. |
Related to revisiting time/how often can you get image on ground. Orbiting time. |
|
|
What are the two types of scanning sensor systems? |
Across-track (Whiskbroom) Scanning Along-track (Pushbroom Scanning) |
|
|
Describe across-track scanning. |
Uses rotating mirror. Scans lines perpendicular to flight lines. Field of view is perfect circle straight below, distorted oval as sensor scans farther away from center (issues with off-nadir view). Images overlap. Collects range of wavelengths with energy separation system. Larger IFOV-lower spatial resolution. Increase spatial band width-lower spectral resolution. |
|
|
Describe along-track scanning. |
Builds 2D image by recording successive 1D scan lines. Linear array on sensor parallel to flight lines. Multiple arrays for multiple bands. Set by sampling rate. Requires more calibration. Can't put all arrays on sensor, only tests some bands. Longer dwell time over ground cell, better spatial and radiometric resolution. Less moving parts. |
|
|
Describe satellite orbits. |
Elliptical orbit. Earth at one foci. Orbital elements include altitude, inclination, equatorial crossing time, period. |
|
|
Describe tangential-scale distortion. |
IFOV scans at constant angular velocity, doesn't equal constant speed over terrain. If x-axis is direction of flight line and y-axis is across flight line, y coordinates are distorted, need to compute corrected distances. |
|
|
What are the two radiometric calibration methods? |
Internal blackbody reference Air-to-ground correlation |
|
|
Describe internal blackbody reference. |
Sensor includes internal blackbodies, one hot and one cold. Identifies total temp range. Doesn't involve atmosphere. Much more cost-effective. |
|
|
Describe air-to-ground correlation. |
Records temp of surface points and tries to calibrate to atmosphere. More costly. |
|
|
Describe Landsat-7. |
1st satellite for earth resource observation. NASA and US Dept of Interior. Started in 1970s. Currently have Landsat 8. 8 bands Spectral: Panchromatic (0.5-0.9 um); blue,green,red/near-IR)
Spatial: Pan (15 m) Multi (30 m): 5 bands 3 in visible; 1 in near-IR; 2 in mid-IR Thermal (60 m): 1 band in Thermal-IR Pixel on ground: 30 m No 3D imaging, takes photo straight-down. Orbit time is every 16 days (depending on atmospheric conditions)
|
|
|
What is SPOT? |
French's Landsat. sun synchronous, across-track. Can get 3D look at pt on ground w/ different perspectives. Little better spatial resolution. Have to pay for SPOT. Wider swath widths. Vegetation sensor revisits same pt on ground on a daily basis. |
|
|
Describe the difference between spectral and spatial resolutions. |
To improve spatial resolution, you must reduce the spectral resolution in order to be able to capture the same amount of energy. To improve spectral resolution, you must do the same. |
|
|
Describe the differences and relationship between a 2-bit and 4-bit sensor. |
2^2 and 2^4 is how many grey scales (B/W and other shades of grey) This is the number of values the sensor can capture. The higher the bit depth, the better the radiometric resolution. |
|
|
If you wanted to study the vegetation on a daily basis, would you choose the Landsat or MODIS and why? |
MODIS b/c it has vegetation sensor that has daily global coverage, faster temporal scale than Landsat. |
|
|
Explain why no information could be required by a satellite sensor at a particular wavelength. |
Sensors don't capture energy at all wavelengths. And even if they can, atmospheric conditions may be to blame that could block energy. Water, CO2, and O3 can affect energy transmission to the atmosphere. |
|
|
Why if one photo could a lake appear dark in color and another appear light? |
In the light lake pic, the sensor was faced in the direction facing the reflected sunlight energy off the water. In 1st photo, sensor isn't facing that direction. |
|
|
Describe two ways to measure scale. |
Equations on equation sheet. S=f/H (f=focal length; H=flying height above terrain) S=d/D= photo distance/ground distance can pick 2 points |
|
|
T or F. Image parallax is mostly a defect of sensor lenses. |
False |
|
|
T or F. In general, vegetation is easier distinguished from soil in green band rather than the infrared band. |
False |
|
|
What is the smallest pixel size for satellite imagery considering government constraints? |
0.5 m |
|
|
Discuss advantages/disadvantages of radar. |
Advantages -Can see through clouds -Multiple returns, stronger signals Disadvantages -Complicated -Need to hire specialist -Multiple returns ->lots of energy, can cause bright spots |
|
|
Side-looking radar systems |
Radar sent from aircraft->return signal from object. The taller the object, the higher the pulse strength. Slant Range = 1/2 ct |
|
|
Describe beamwidth in radar. |
Proportional to wavelength of pulse. Long antenna = small beamwidth |
|
|
What's synthetic aperture radar?
|
Using synthesized antenna length(short) through recording + processing. Use sensor motion along track. Mathematically transform antenna into array of antennas. Relies on detecting Doppler shifts (Points forward -> high frequency; Points behind->low frequency). Issues: resolution is related to antenna length (res = 1/2 antenna length); Tradeoffs (operating range, wavelength, system complexity, resolution, antenna size) |
|
|
What scale distortions exist? |
Slant-range recording Ground-range recording |
|
|
Discuss scale variation in slant range. |
Across track related to speed of light
Along track related to speed of aircraft |
|
|
Discuss parallax. |
What is it? -Apparent shift of object when viewed at different angles -Affects closer objects more severely How is it used? -Generates 3D scene from multiple position images -Aerial photography |
|
|
Discuss transmission |
-Factors: wavelength, polarization -Major effects on wavelengths < 4 cm Radar signal unaffected by clouds Can detect echoes from heavy precip. |
|
|
Polarization |
Wave filtering -Can use different modes of polarization -Restrict wave vibrations to single plane perpendicular to propagation direction Restrict transmission -Horizontal -Vertical |
|
|
Discuss radar image speckle. |
"Salt and pepper" Causes: -Multiple scattering returns occur within each cell -Random constructive/destructive interference How to deal: -Reduce through image processing -Cannot completely remove |
|
|
Why is the role of illumination in stereo radar images? |
SAR images have brightness gradient. -Ground resolution cell smaller at near range -Backscatter varies with local incident angle |
|
|
What is digital image processing? |
Manipulation and interpretation of digital images |
|
|
What are the different types of image processing? |
-Image rectification and restoration -Image enhancement -Image classification -Hyperspectral analysis -Biophysical monitoring |
|
|
What is the difference between supervised/unsupervised classification? |
Unsupervised: pick # and press button. accuracy, time + cost. Supervised: pick points and say cover type. Computer will try to generalize |
|
|
Image rectification and restoration |
-Correct distorted/degraded images -Geometric or radiometric distortion -Procedure depends on sensor -Considered as preprocessing |
|
|
Image enhancement |
-Display/record data more effectively -Often aims to increase visual distinction between features |
|
|
Image classification |
-Use quantitative techniques to identify features in scene -Spectral pattern recognition -Output goal: thematic map e.g. land cover |
|
|
Hyperspectral analysis |
-Many multispectral methods apply to hyperspectral data -Differences: amount of data, data correlation |
|
|
Biophysical monitoring |
-Relate digital data to phenomena measured on ground |
|
|
What is image resampling? |
Deriving pixel values for a new image from an existing image. Every pixel in image has DN value. |
|
|
What methods of image resampling are there? |
Nearest-neighbor (replacing every pixel with # of pixels of same color) Bi-linear interpolation (distance weighted average of nearest 4 pixels) Cubic convolution (distance weighted average of nearest 16 pixels) |
|
|
How do you correct for differences in illumination?
|
Sun position determines illumination.
Earth-Sun Distance Correlation. Irradiance descreases as square of earth-sun distance increases E = E0 cos () / d^2 Atmospheric Effects -Atmosphere attenuates signal-reduces energy -Scatters energy toward sensor L (radiance of sensor)= pET/pi + Lp p = reflectance E = irradiance of object T = transmission Lp= path radiance |
|
|
How do you convert spectral radiance into a digital number? |
A-to-D process converts radiance to DN Digital sensors designed to have linear response DN = G x L + B G(slope): L (spectral radiance); B L = (Lmax - Lmin/255) DN + Lmin |
|
|
What is contrast manipulation? |
Contrast-enhancement parameters convert the digital input data to an image with appropriate density and contrast |
|
|
What methods can be used for contrast manipulation? |
Grey-level tresholding Level slicing Contrast stretching |
|
|
What's grey-level tresholding? |
Segment image into two classes: one having pixel values below gray level and one for values above |
|
|
What's level slicing? |
Combining DNs of different values within a specified range into a single value |
|
|
What's contrast stretching? |
Stretch DN values to maximize range DN' = (DN - Min/Max - Min)255 |
|
|
What's spatial filtering? |
Emphasize/deemphasize changes in spatial frequency Rapid change->High spatial frequency->rough Slow change->Low spatial frequency->smooth |
|
|
How is convolution applied? |
Pass moving window over image Apply operators to image->generate output |
|
|
What special purpose filters exist for convolution? |
Smoothing Filter (Average) Smoothing Filter (Majority) Edge Enhancing Filter: -1 -1 -1 -1 9 -1 -1 -1 -1 |
|
|
Describe supervised classification. |
-Requires input from image analyst -Specify regions representing cover types -Develop rules from multispectral input -Assign every pixel in image a class based on rules |
|
|
What are the approaches of supervised classifcation? |
Miniumum-distance-to-means Parallelepiped Maximum likelihood Training Development |
|
|
Describe minimum-distance-to-means |
-Calculate mean vector for each class -Calculate distance from unknown pt to each class mean -Assign class w/ min distance Simple and efficient Issues: -No consideration of degree of variability of class -Establish treshold distance |
|
|
Describe parallelepiped. |
Consider range of values within each class -With 2 bands (2-D) range->rectangle -With n bands (n-D) range->parallelepiped -Consider stepped decision region Issue: data correlation |
|
|
Describe maximum likelihood classifier |
-Evaluate variance + covariance of spectral patterns -Consider statistical probability of pt belonging to class Issues: -Assumes normally distributed pt data -Limited implementation -Need large # of computations in classifying -Improved processors->larger images |
|
|
What are some issues with supervised classification? |
-Need to have knowledge of ground area
-Challenging to separate spectrally similar classes -Iterative process -Can classify using subclasses + merge at end |
|
|
What are the types of spectral classification? |
Supervised Unsupervised Hybrid |
|
|
Discuss unsupervised classification.
|
-No training data used -Aggregate pixels based on natural clusters -Cluster on spectral similarities only |
|
|
Contrast supervised vs unsupervised classification. |
Supervised: define classes->consider separability Unsupervised: consider separability -> define classes |
|
|
Name one advantage and disadvantage of unsupervised classification. |
Advantage: No training data Disadvantage: Rare classes not well represented |
|
|
What is the motivation behind hybrid classification? |
It's the best of both supervised and unsupervised. 1)Unsupervised of training area of each class 2) Make subclasses 3) Examine, edit if necessary 4) Supervised of all subclasses 5) Put subclasses to desired classes |
|
|
Name an implementation example of hybrid classification. |
Guided clustering |
|
|
Why do we do spectral mixture analysis? |
-Determine "pure" response of known cover types -Compare mixed responses to combo of "pure" response |
|
|
What does the accuracy matrix show? |
Provide class-by-class comparison of reference and classified data |
|
|
How do you determine overall accuracy? |
Percentage of pixels correctly classified Sum of diagonals, divide by diagonal total sum |
|
|
How do you determine producer's accuracy? |
Per class indicator of how well reference data was classified Take element for column divide by column total sum |
|
|
How do you determine user's accuracy? |
Per class indicator of correctness of classified data Take element for row divide by row total sum |
|
|
What are the cons/pros of random/stratified sampling? |
Pro: Provides unbiased assessment Cons: -Data collection may be difficult + $$$ -Registration between classified and reference layers critical -SRS undersamples infrequent cover types |
|
|
Discuss hyperspectral sensors. |
-Called "imaging spectrometers" -Used to construct reflectance curves -Many narrow spectral bands -Visible, NIR, MIR, and TIR -Generally contiguous bands -Bands 10 nm wide, 200+ bands |
|
|
Discuss the hyperspectral cube. |
KNOW It |
|
|
Name one major advantage of hyperspectral analysis vs. multispectral. |
Multispectral->separate objects but Hyperspectral->identify/characterize objects |
|
|
Name one major disadvantage of hyperspectral analysis vs multispectral. |
Multispectral: collects isolated, broad bands (more energy) Hyperspectral: Lots of narrow bands->less total energy |
|
|
Discuss spectral angle mapping. |
-Method for comparing image/reference spectra -Consider observed response as n-dimensional vector -Change in scene illumination: changes length of vector but not direction |
|
|
LIDAR |
-vertical penetration (vertical graph) -peaks show top of canopy -Processing can be time-consuming Vertical resolution 1st (top) + last (ground) can calculate height -Highest cost: plane flying time -Can decrease costs by minimizing overlaps (distortion), sensor with higher FOV (don't need as many scanning lines) |
|
|
T or F: Spectral resolution is defined as the number and width of wavelength intervals that can be detected by sensors. |
True |
|
|
For an across-track scanning system, a large IFOV is desirable to record fine spatial detail. |
False |
|
|
___-track scanning system scans perpendicular to the direction of movement of the sensor |
Across |
|
|
Water typically has a much ___ tone in black and white IR photos because sunlight reflection from water in the near-IR is considerably ___ than in the visible part of the spectrum.
|
darker, less |
|
|
A Landsat TM image consists of seven spectral bands. If the bands 2(green), 3 (red), and 4 (near-IR) are assigned to respectively B, G, and R colors for display, green vegetation appears ___ because of its high reflectance in the ___ band. |
red, near-IR |
|
|
Name some advantages of Along-track. |
-Longer dwell time->stronger signal recorded->better radiometric resolution -Geometric integrity is better b/c less moving parts -Higher reliability and longer life expectancy |
|
|
T or F: Landsat 7 uses a pushbroom sensor. |
False |
|
|
T or F: A scanning mirror rotates at constant angular velocity, hence a constant ground resolution cell size over the entire across-track imagery. |
False |
|
|
T or F: Across-track imagery presents 1-D relief displacement in the cross-track direction. Image scale in the direction of flight is essentially constant. |
True |
|
|
Advantages of across-track? |
-Presents 1D relief displacement in cross-track direction. Image scale in direction of flight is constant |
|
|
T or F: Atmospheric effects on radiant temperature measurements from thermal images usually cannot be ignored. |
True |
|
|
T or F: Internal blackbody reference and air-to-ground correlation are the two steps to radiometrically calibrate thermal images. Both take into consideration the impact of atmospheric conditions. |
False |
|
|
__ resolution is defined as the revisiting frequency at which images are captured over a specific area. |
Temporal |
|
|
A across-track scanner uses a ____ to separate thermal and non-thermal wavelengths. |
dichroic grating |
|
|
__ is defined as the ratio of radiation emitted by an object at a given temperature to the radiation emitted by a blackbody at the same temperature. |
Emissivity |
|
|
T or F: SPOT can create 3-D images due to its off-nadir viewing capability.
|
True |
|
|
T or F: The Landsat satellites are operated in a sun-synchronous orbit, and the SPOT satellites have a geostationary orbit. |
False |
|
|
T or F: The Scan Line Corrector (SLC) in the ETM+ instrument on Landsat-7 failed in 2003. It is used to compensate for the forward motion of the spacecraft so that scan lines are aligned parallel to each other. |
True |
|
|
T or F: Radar is an active microwave system that detects sunlights reflected by terrain features. |
False |
|
|
T or F: Radar images display ranges, or distances, from terrain features in the antenna. |
True |
|
|
T or F: The azimuth resolution is solely determined by the angular beamwidth of the antenna; thus it remains constant at all points for a specific system. |
False |
|
|
Due to the ___ used, microwaves are capable of penetrating most atmosphere conditions, e.g. haze, smoke and clouds. |
long wavelengths |
|
|
T or F: The overall objective of image classification is to automatically categorize pixels in an image into a predefined label. |
True |
|
|
T or F: The quality of unsupervised classification has no relationship with the quality of the training data, but completely depends on the classification method. |
False |
|
|
T or F: In a supervised approach, image pixels are first automatically aggregated based on spectral similarities, and then an analyst attempts to label the classes. |
False |
|
|
What is the major difference between supervised and unsupervised classification? |
Unlike supervised classification, no initial training data is used in unsupervised classification. |
|
|
T/F: An averaging filter is usually used for post-classification smoothing to eliminate isolated pixels on the classified image. |
False |
|
|
T/F: It is better to use training data instead of testing data for accuracy assessment due to time or cost constraints. |
False |
|
|
T/F: Simple random sampling tends to undersample rare but potentially important land cover types. |
True |
|
|
T/F: Hyperspectral sensors acquire images in many, very narrow spectral bands. |
True |
|
|
T/F: The visible portion of the EM Spectrum used in remote sensing has longer wavelengths than the infrared portion. |
False |
|
|
T/F: Reflectance is defined as the fraction of incident energy reflected by an object. |
True |
|
|
What makes the sky blue is the ____ sunlight by particles in atmosphere. |
scattered |
|
|
T/F: For human eyes it is easier to detect remotely warmer than cooler cycles. |
True |
|
|
T/F: Spatial resolution is the minimum distance between two adjacent objects that can be distinguished by a remote sensing system. |
True |
|
|
The digital number (DN) represents the average ____ measured in each pixel. |
radiance (or brightness) |
|
|
T/F: Vertical photographs are those acquired with camera axis truly vertical at the moment of exposure, while oblique photographs are those taken with an intentional inclination of the camera axis. |
True |
|
|
t/F : Photographs taken over terrain of varying elevation will exhibit a uniform scale. |
False |
|
|
T/F: Due to the nature of parallax, it is possible to measure feature heights and create a 3-D viewing of overlapping photographs from a stereo pair. |
True |
|
|
T/F: Under similar conditions, high altitude photography of an area exhibits less relief displacement than low altitude photography of the same area. |
True |
|
|
T/F: There is no relief displacement at the principal point if the image is truly vertical. |
True |
|
|
T/F: Compared to single-lens frame camera, panoramic cameras cover a much larger area but yield more apparent geometric distortions on the photographs. |
True |
|
|
The _____ has spectral sensitivity at longer wavelengths. |
infrared B&W film |
