Radar images
Radar images radar image DNs linked to backscattered field Backscattered field depends on wave-target interaction, with different factors relevant to it: within-pixel coherent combination (speckle noise) surface roughness; surface orientation wrt radar; electrical properties and humidity content of the target
speckle noise speckle noise is the result of a noise-like process degrading the quality of data recorded by a coherent system speckle is due to unforeseeable (noise-like) within-pixel interference, generating random amplification or damping of the backscattered field speckle visually turns into salt-and-pepper noise
Within-pixel coherent sum + +
Visual appearance Speckle noise and the rough look it gives to images is typical of radar data This is a clue to tell optical images from radar images
de-speckling Speckle reduction is obtained by de-speckle filtering and by multi-looking. "multi-looking different doppler bands are averaged in SAR de-speckle filtering: specialized filters exist to reduce noise La riduzione di questi effetti è ottenuta al prezzo di una riduzione della risoluzione spaziale.
De-speckling http://remotesensing.spiedigitallibrary.org/data/journals/appres/24182/053502_1_14.png
Multiplicative model Multiplicative model for speckle noise: I = 2σ 2 n The signal is a product of: Information contribution: Random contribution: p(i σ) = 1 P(n σ) = e n = P(n) I 2σ 2 e 2σ 2 = p(i)
Roughness Surface roughness is an important factor in determining backscatter: smooth surfaces mirror-like reflect EM waves away from sensor rough surfaces scatter EM waves including towards the sensor Roughness is relative to λ. Average variations for rough surfaces are greater than half the projection of λ on the surface normal
Roughness
Corner reflection Strong reflection, typical for man-made structures
Roughness and corner reflection
Corner reflectors Corner reflection is sometimes generated for purposes of calibration using corner cube structures http://uavsar.jpl.nasa.gov/rosamond.html http://en.wikipedia.org/wiki/corner_reflector
Corner reflectors Corner reflectors cause strongest backscatters image from: SANDIA REPORT SAND2008-0396 Unlimited Release Printed January 2008 Reflectors for SAR Performance Testing Armin W. Doerry
Local incidence angle Surfaces with orientation nearer to the wave direction generally backscatter more strongly http://www.geog.ucsb.edu/~jeff/115a/remote_sensing/radar/radar2.html
Conductivity Conductive matter reflects strongly:
Volume scattering vs. λ
Volume scattering Volume scattering is connected to multiple reflection and scattering in a inhomogeneous means: vegetation dry soil, sand ice Need to separate volume scattering from surface scattering contributions (not always easy)
Humidity The presence of humidity alters the backscattering properties of a target More humid targets tend to reflect more from their surface and less from their volume In the case of vegetation, longer wavelengths penetrate deeper
SAR Interferometry The availability of more than one datum allows extracting 3-D information It can be extracted from range difference (radargrammetry) phase difference (interferometry)
Radargrammetry z S 1 B α S 2 r 2 h 0 ϑ r 1 h y
Interferometry z Φ 1 S 1 h 0 B ϑ α S 2 r 1 Φ 2 r 2 Φ 1,2 = 2ω c r target 1,2 + Φ 1,2 Φ 2 target Φ 1 target (target seen from same angle) h y
Interferogram Interferogram: 0 2π angle cycles are represented by the iris colour cycles ( fringes )
Baseline 12 metres 60 metres Longer baselines narrower fringes more height precision, but also more height ambiguity
Phase unwrapping the interferometric phase must be unwrapped to a continuous angle difference before being translated into height Complex and error-prone task
Vesuvius DEM example
along-track interferometry tandem satellites two antennas, along-track across-track interferometry two orbits antenna mast differentiale interferometry (DInSAR) interferometry between-pairs
Interferometric coherence Normalized complex crosscorrelation between the two images in the interferometric pair computed on a sliding kernel γ = E( s s *) 1 2 E( s s *) E( s s *) 1 1 2 2
Strong reflector If a pixel contains a strong reflector: speckle noise becomes less effective (hypothesis not verified) the phase of the reflector can be estimated if many samples available A long, multitemporal series can allow estimation of phase drift along time line-of-sight motion of targets Precision can reach very fine levels (few mm/year) This technique is called Permanent Scatterers : Alessandro Ferretti, Claudio Prati, and Fabio Rocca: Permanent Scatterers in SAR Interferometry, IEEE TGRS, Vol. 39, No. 1, pp. 8-20, January 2001
Example From the Telerilevamento Europa web site:
Used mainly for local, precise, continuous monitoring of a specific site (es: potential landslides sites)