lecture 4 - app. rs

Upload: mario-ultimateaddiction-hylton

Post on 08-Apr-2018

222 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/7/2019 Lecture 4 - App. RS

    1/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 1 1

    The process of manipulation and interpretationof digital images using a computer;Began in the 1960s with airborne multispectralscanner data and digitized aerial photos;

    Digital image data became available after 1972with the launching of Landsat-1;Typically include:- Radiometric preprocessing adjusting

    digital values for the effect of scattering andhaze; and

    - Geometric preprocessing registeringimage with a map or other image;

    Lecture 4Digital Image Processing

  • 8/7/2019 Lecture 4 - App. RS

    2/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 2 2

    Categories of Preprocessing:

    Feature or Information Extraction:Refers to statistical characteristic of theimage e.g. individual bands or combination of band values;Discarding of data containing noise anderrors, isolating only components that areessential to portraying essential elements;Increases accuracy;Multispectral data by nature consist of several channels (3, 4, or 7);

  • 8/7/2019 Lecture 4 - App. RS

    3/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 3 3

    Reducing the number of spectral channel, or bands being analyzed reducingcomputational demand;After completion of feature selection, analysisworks with fewer and more effective bands increasing speed and reduces cost of analysis;Uses the computer to recognize and classifypixels or neighborhood of pixels on the basisof their spectral-radiometric response (DNs);

  • 8/7/2019 Lecture 4 - App. RS

    4/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 4 4

    S ubsets:

    Selection of a portion of a large image to showonly the region of interest:Must be registered to other data using landmarks matching images to maps or other

    maps;Should be large enough to provide sufficientnumber of training fields for image classificationor sites for accuracy assessment;

  • 8/7/2019 Lecture 4 - App. RS

    5/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 5 5

    R adiometric Correction;

    Radiometric data allows the analysis to relatebrightness recorded by aerial sensors tocorresponding brightness of ground feature;Radiance measured is influenced by:

    - changes in scene illumination,- atmospheric condition,- viewing geometry (greater in airborne data

    collection), and

    - instrument response;

  • 8/7/2019 Lecture 4 - App. RS

    6/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 6 6

    A tmospheric Correction Procedure:

    Atmospheric correction is based on examination of reflectance from objects of known or assumedbrightness;Using the principles of atmospheric scattering(related to wavelength, size of particles, andabundance);Employ the use of features of known brightness e.g.water bodies or shadows cast by clouds or largetopographic features;

    Brightness value at or near zero (strong absorption)in the IR portion of the spectrum and very little IRenergy is scattered;

  • 8/7/2019 Lecture 4 - App. RS

    7/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 7 7

    Effect of atmospheric scatteringand absorption on spectral-raidometric responses

  • 8/7/2019 Lecture 4 - App. RS

    8/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 8 8

    Each cell (or pixel) has anumerical value which is afunction of the material presentwithin that pixel (i.e. the

    reflective or emissive propertiesof the material) and the spectralregion by which the sensor isrecording.

    S atellite Images A re Digital Data

  • 8/7/2019 Lecture 4 - App. RS

    9/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 9 9

    T opographic effect:Slope correction (normalization) is the removalof all topographically induced illumination Objects showing the same brightness value

    despite their different orientation to the sunsposition;

    Commonly require a DEM of the study area for slope-aspect topographic correction;

  • 8/7/2019 Lecture 4 - App. RS

    10/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 10 10

    The amount of illumination is a function of the angle that light hits the slope;S un elevation correction accounting for elevation correction and for seasonalposition of the sun relative to the earth;

    - Images acquired under different solar illumination angles are normalized bycalculating pixel brightness values ,assuming the sun was at the zenith (90 0

    minus the solar elevation angle) on eachdate;

  • 8/7/2019 Lecture 4 - App. RS

    11/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 11 11

  • 8/7/2019 Lecture 4 - App. RS

    12/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 12 12

    G eometric correction:All RS imagery is inherently subject togeometric distortions;Geometric corrections are intended tocompensate for these distortions so that thegeometric representation of the imagery will beas close as possible to the real world.Source of distortions in images include:- variation in altitude;- velocity of the sensor platform;- panoramic distortion;- earth curvature;

    - atmospheric refraction;- relief displacement; and- nonlinearities in the sweep of a sensors

    IFOV;

  • 8/7/2019 Lecture 4 - App. RS

    13/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 13 13

    Variations are either systematic, or predictablein nature and can be accounted for by accuratemodeling of the sensor and platform motion andthe geometric relationship of the platform withthe Earth.3 R s of G eometric Correction:

    R ectification: Transformation of a geometricallydistorted image so that it can be registered to amap moving pixels to a correct map location;R esampling: Determination of DN values to fill

    in the output matrix of the rectified image;R egistration: Overlaying two or more images or maps so that they coincide geometrically;

  • 8/7/2019 Lecture 4 - App. RS

    14/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 14 14

    G eometric correction - R esampling:

    Corrected by analyzing well-distributed groundcontrol points (GCPs);- i.e. known ground locations that can be located

    on the digital image;

    Identifying image coordinates (i.e. row, column)in the distorted image and matching them to their true position in ground coordinates (e.g. latitude,longitude);

    Geometric transformation is performed using anaffine transformation (preserve lines andparallelism);

  • 8/7/2019 Lecture 4 - App. RS

    15/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 15 15

    Transformation equations are fitted using least-square regression equation to determinecoefficients interrelating the corrected mapcoordinate to the distorted image coordinates;The resampling process calculates new pixelvalues from the original digital pixel value;Three common methods of resampling:N earest neighbor:Assign each corrected pixel the value from thenearest uncorrected pixel (labeled a in

    diagram);Adv. simple and preservation of original valueDisadv. Creates positional error

  • 8/7/2019 Lecture 4 - App. RS

    16/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 16 16

    B ilinear interpolation:

    Value of each output pixel based on the distance-weighted average of the four nearest input pixel (aand b in diagram);Image smoother in appearance than nearest

    neighbor;Disadv. - Brightness values in input image are lost;- Decreases spatial resolution (smearing) causedby averaging small features with adjacent

    background pixels;

  • 8/7/2019 Lecture 4 - App. RS

    17/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 17 17

    Cubic convolution:

    W eighted average value of adjacent 16 pixels (a,b, and c in diagram);Appearance sharper, computation more intense,require a larger number of GCPs;

  • 8/7/2019 Lecture 4 - App. RS

    18/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 18 18

  • 8/7/2019 Lecture 4 - App. RS

    19/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 19 19

    Advantages of Digital Image Processing ::Cost-effective for large geographic areas;

    Cost-effective for repetitive interpretations;Cost-effective for standard image formats;

    Consistent results;Simultaneous interpretations of severalchannels;Complex interpretation algorithms possible;

    Speed may be an advantage;Explore alternatives;Compatible with other digital data ; ;

  • 8/7/2019 Lecture 4 - App. RS

    20/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 20 20

    Disadvantages of Digital Image Processing ::

    Expensive for small areas;Expensive for one-time interpretations;Start-up costs may be high;Requires elaborate, single-purposeequipment;Accuracy may be difficult to evaluate;Requires standard image formats;

    Data may be expensive, or not available;Preprocessing may be required;May require large support staff;

  • 8/7/2019 Lecture 4 - App. RS

    21/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 21 21

    Digital vs. Manual Interpretation:

    M A NU AL I N T E R P R E TAT IO N Traditional and intuitive;Simple, inexpensive equipment;Uses brightness and Spatial content of the

    image;Usually single channel data or three channels atmost;

    Subjective, concrete, qualitative

  • 8/7/2019 Lecture 4 - App. RS

    22/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 22 22

    DI G I TAL I N T E R P R E TAT IO N

    Recent application;Requires specialized trainingComplex, expensive equipmentsRelies chiefly upon brightness and spectralcontent, limited spatial;Frequent use of data from several channels;Objective, abstract, quantitative;

  • 8/7/2019 Lecture 4 - App. RS

    23/23

    1/2/051/2/05 Applied Remote SensingApplied Remote Sensing 23 23

    A ssigned R eading L illesand and Kiefer. 1994. R emote S ensing and

    Image Interpretation. Chap. 7

    Campbell. Introduction to R emote S ensing. Chap.10