1 / 28

Remote Sensing and Image Processing: 7

Remote Sensing and Image Processing: 7. Dr. Mathias (Mat) Disney UCL Geography Office: 301, 3rd Floor, Chandler House Tel: 7670 4290 (x24290) Email: mdisney@geog.ucl.ac.uk www.geog.ucl.ac.uk/~mdisney. Today…. Optical sensors Spatial and spectral resolutions

alban
Télécharger la présentation

Remote Sensing and Image Processing: 7

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Remote Sensing and Image Processing: 7 Dr. Mathias (Mat) Disney UCL Geography Office: 301, 3rd Floor, Chandler House Tel: 7670 4290 (x24290) Email: mdisney@geog.ucl.ac.uk www.geog.ucl.ac.uk/~mdisney

  2. Today….. • Optical sensors • Spatial and spectral resolutions • Choices we make for different applications • Trade-offs of coverage against detail

  3. Resolution • What do we mean by “resolution” in RS context • OED: the effect of an optical instrument in making the separate parts of an object distinguishable by the eye. Now more widely, the act, process, or capability of rendering distinguishable the component parts of an object or closely adjacent optical or photographic images, or of separating measurements of similar magnitude of any quantity in space or time; also, the smallest quantity which is measurable by such a process.

  4. Resolution • Even more broadly • Not just spatial.... • Ability to separate other properties pertinent to RS • Spectral resolution • location, width and sensitivity of chosen bands • Temporal resolution • time between observations • Radiometric resolution • precision of observations (NOT accuracy!)

  5. Shrink by factor of 8 Spatial resolution • Ability to separate objects in x,y

  6. Spatial resolution • Smallest object we can determine on surface • Ranges from < 50cm to > 5km • Function of altitude of sensor….. • Further away we are, lower resolution for fixed system • ….and optics of instrument • More powerful the telescope we use, more detail we see BUT smaller area we can cover • So tradeoff detail (high spatial resolution) v coverage (lower spatial resolution)

  7. Spatial resolution • Spatial resolution • formal definiton: a measure of smallest angular or linear separation between two objects that can be resolved by sensor • Determined in large part by Instantaneous Field of View (IFOV) • IFOV is angular cone of visibility of the sensor (A) • determines area seen from a given altitude at a given time (B) • Area viewed is IFOV * altitude (C) • Known as ground resolution cell (GRC) or element (GRE)

  8. IFOV and ground resolution • Image pixels often idealised as rectangular array with no overlap • In practice (e.g. MODIS) • IFOV not rectangular • function of swath width, detector design and scanning mechanism • see later.... MODIS home page: http://modis.gsfc.nasa.gov/

  9. AVHRR IFOV • Scan of AVHRR (Advanced Very High Resolution Radiometer) • elliptical IFOV, increasing eccentricity with scan angle

  10. Aside: what’s in a pixel? • mixed pixel (mixel) problem in discrete representation Cracknell, A. P. (1998) Synergy in remote sensing: what’s in a pixel?, Int. Journ. Rem. Sens., 19(11), 2025-2047

  11. So.....? • If we want to use RS data for anything other than qualitative analysis (pretty pictures) need to know • sensor spatial characteristics • sensor response (spectral, geometric)

  12. Examples • High (10s m to < m) • Moderate (10s - 100s) • Low (km and beyond) Jensen, table 1-3, p13.

  13. Low v high spatial resolution? • What is advantage of low resolution? • Can cover wider area • High res. gives more detail BUT may be too much data • Earth’s surface ~ 500x106 km2 ~ 500x106 km2 • At 10m resolution 5x1012 pixels (> 5x106 MB per band, min.!) • At 1km, 500MB per band per scene minimum - manageable (ish) • On the other hand if interested in specific region • urban planning or crop yields per field, • 1km pixels no good, need few m, but only small area • Tradeoff of coverage v detail (and data volume) From http://modis.gsfc.nasa.gov/about/specs.html

  14. Spectral resolution • Measure of wavelength discrimination • Measure of smallest spectral separation we can measured • Determined by sensor design • detectors: CCD semi-conductor arrays • Different materials different response at different  • e.g. AVHRR has 4 different CCD arrays for 4 bands • In turn determined by sensor application • visible, SWIR, IR, thermal??

  15. Tradeoffs • Notice how concept of tradeoff keeps cropping up • We almost always have to achieve compromise between greater detail (spatial, spectral, temporal, angular etc) and range of coverage • Can’t cover globe at 1cm resolution!! • Resolution determined by application (and limitations of sensor design, cost etc.)

  16. Recap: continuous spectrum • Where do we look?? • Remember atmospheric windows!

  17. Ideal bandpass function Spectral resolution • Characterised by full width at half-maximum (FWHM) response • bandwidth > 100nm • but use FWHM to characterise: • 100nm in this case From: Jensen, J. (2000) Remote sensing: an earth resources perspective, Prentice Hall.

  18. Spectral information: vegetation vegetation

  19. Ch1: 0.58-0.68m Ch2: 0.73-1.1 m Ch3: 1.58-1.64 m Ch4,5: 10.5-11.5 & 11.5 - 12.5 m Broadband & narrowband • AVHRR 4 channels, 2 vis/NIR, 2 thermal • broad bands hence less spectral detail From http://modis.gsfc.nasa.gov/about/specs.html

  20. Broadband & narrowband • CHRIS-PROBA • different choice • for water applications • coastal zone colour studies • phytoplankton blooms From http://www.chris-proba.org.uk

  21. Multispectral concept • Measure in several (many) parts of spectrum • Exploit physical properties of spectral reflectance (vis, IR) • emissivity (thermal) to discriminate cover types From http://www.cossa.csiro.au/hswww/Overview.htm

  22. Multispectral concept • MODIS: 36 bands, but not contiguous • Spatial Resolution: 250 m (bands 1-2), 500 m (bands 3-7), 1000 m (bands 8-36) • Why the difference across bands?? • bbody curves for reflected (vis/NIR) & emitted (thermal) From http://modis.gsfc.nasa.gov/about/specs.html

  23. MODIS: fires over Sumatra, Feb 2002 • Use thermal bands to pick fire hotspots • brightness temperature much higher than surrounding From http://visibleearth.nasa.gov/cgi-bin/viewrecord?12163

  24. x z y Multi/hyperspectral • Multispectral: more than one band • Hyperspectral: usually > 16 contiguous bands • x,y for pixel location, “z” is  • e.g. AVIRIS “data cube” of 224 bands • AVIRIS (Airborne Visible and IR Imaging Spectroradiometer) From http://aviris.jpl.nasa.gov/ & http://www.cossa.csiro.au/hswww/Overview.htm

  25. Examples • Some panchromatic (single broad bands) • Many multispectral • A few hyperspectral Jensen, table 1-3, p13.

  26. Broadband v narrowband? • What is advantage of broadband? • Collecting radiation across broader range of  per band, so more photons, so more energy • Narrow bands give more spectral detail BUT less energy, so lower signal • More bands = more information to store, transmit and process • BUT more bands enables discrimination of more spectral detail • Trade-off again

  27. Recap • Spatial resolution • IFOV, FOV, GRE and PSF • Spectral resolution • Choice of bands and bandwidth • Tradeoffs • Higher resolution means more detail, but more data • Also higher resolution means lower energy i.e. needs more sensitive detectors

  28. Practical • Assessed practical • Supervised classification of Churn Farm image • Set up training data (choose regions of interest) ROIs • Look at class extents in feature space • b1 v b2, b1 v b3, b2 v b3 • are classes too broad in which case change / redo them? • Try various classifications • Accuracy? • Maybe try unsupervised?

More Related