ANALYSING BACKSCATTER RESPONSE AND S C A T T E R I N G B E H A V I O U R O F L A N D F E A T U R ES U S I N G F R E E M A N D U R D E N D E C O M P O S I T I O N , V O L UM E SCATTERING INDEX & SPAN [TOTAL B A C K S C A T T E RP O WE R / I N T E N S I T Y ] TEJ D. CHAVDA (PG191052) M.TECH GEOMATICS (2ND SEMESTER), FACULTY OF TECHNOLOGY, CEPT UNIVERSITY, AHMEDABAD, GUJARAT
Abstract Synthetic aperture radar (SAR) is an active form of remote sensing that utilizes radar signals to capture high resolution images and is similar to optical remote sensing. SAR is a coherent airborne or space borne sidelooking radar system which simulates an extremely large antenna or aperture electronically, and this generates high-resolution remote sensing imagery. Coherence describes the phase relationships between the transmitted and the received pulses. The current study is an attempt to analyze backscatter response and scattering behavior of various land-use land cover feature classes in Vancouver seaport city, Canada, located in the Lower Mainland region of British Columbia. It is located between latitudes 49°21' and 49°09'N and longitudes 123°14' and 122°52' W. RADARSAT-2 C-band (Quad-Pol Dataset) has been used for this study and it was acquired on 15 th April, 2008. The overall study is divided into two phases. In the first phase, Single look Complex (SLC) image has been converted into backscatter image [σ° (dB)]. This has been further analyzed and backscatter response and average dB value in HH Polarization for each feature class has been plotted. In the second phase, polarimetric analysis has been carried out for the same SAR image. In polarimetric analysis, polarimetric incoherent decomposition and polarimetric parameters have been studied. For incoherent Decomposition, Freeman-Durden method has been adopted. Polarimetric parameters such as volume scattering index and span index have
been used to further strengthen the analysis and plot the backscatter values for each land feature class. Keywords — Synthetic aperture radar (SAR), Coherence, Sidelooking radar, SAR C-Band data (Quad-pol), Backscatter value in dB, LULC Classification, SAR-Polarimetric Analysis
1. Introduction Radar stands for Radio Detection and Ranging and is defined as an object-detection system. This system basically comprises of an electromagnetic wave transmitter and receiver, usually operating in the range of microwave frequencies. The received radar signal provides information about properties of the target features on the ground. Imaging radar system are the active sensors which do not rely on sun as the energy source. which uses radio waves to determine the range, altitude, direction, or speed of objects. Conventionally, radar allows an operator to detect the presence, direction, and speed of a target, weather body, or terrain object. A single radar image is usually displayed as a grey scale image. Polarization of the radiation is also an important term while discussing the radar systems. Polarization refers to the orientation of the electric field. Studying different polarizations in the given image gives us additional information on the physical properties of the target.
Vertically and horizontally mounted antennas are designed to transmit or receive vertically and horizontally polarized waves, respectively. Therefore, changes in polarization cause changes in the magnitude of the received signal due to the inability of the antenna to receive polarization changes.
Figure 1 Horizontally Polarized Wave (H)
Figure 2 Vertically Polarized Wave (V)
Polarization
2. Study Area Here for the study, we utilize the data set consisting the Greater Vancouver area, Canada. The test site is very diverse in nature consisting a wide variety of features to classify. The area consists of urban settlements including Richmond area, rotated urban areas west to New Westminster. Rugged mountains in northern Vancouver, rivers merging to the Strait of Georgia and crop-lands in the Fraser River Delta RADARSAT-2 data has been acquired on May 2008 over Vancouver area in full polarimetric mode. Ground-truth parameters was also collected synchronous with the satellite pass. Near Range Incidence Angle is 34.49o and Far Range Incidence Angle is 36.08o .The dataset of RADARSAT-2 was acquired in Fine Quad mode with Q15 beam. It has been captured in descending pass direction inferring the snap is recorded on the sunlit side as the orbit of the SAR system is sun-synchronous. The geographic location of the study area lies between 49.2827° N, 123.1207° W, as presented in Figure.
The polarization state of an electromagnetic wave can change when the electromagnetic wave scatters from a target. • HH - for horizontal transmit and horizontal receive, • VV - for vertical transmit and vertical receive, • HV - for horizontal transmit and vertical receive, and • VH - for vertical transmit and horizontal receive.
Figure 3 Vancouver City (Google Earth Image)
Figure 5 Image in HH Polarization
Figure 4 Image in HV Polarization
S.R.NO.
BAND INFORMATION
C-BAND
1
Product type
SLC (Single look complex)
2
Transmitting Polarization
H, V
3
Receiving Polarization
H, V
4
Polarization
HH, HV, VH, VV
5
Acquisition Mode
Fine-Quad Polarization
6
Acquisition Date
15 Apr,2008
7
Mode of Travelling
Descending
8
Look time
Right
9
Operating Frequency
5404 MHz
3. OBJECTIVE OF THIS STUDY
10
Mission
The purpose of this study is to identified various land features and record their backscatter response to analyze the behavior or target features towards microwave radar.
RS2 (RADARSAT2)
11
Average Height
Scattering behavior of land features is studies and understood by using Polarimetric Decomposition analysis and polarimetric parameters. The SAR SLC image has been analyzed in terms of freeman Durden decomposition, volume scattering index & span index (Polarimetric Parameters).
4. DATA & SOFTWARE RADARSAT-2 C-Band (Quad-Pol Dataset) used for this study. Further information given in table as follows: Pre-Processing of an image done by using SNAP software. Backscattering response, scattering behavior and average dB intensity of land feature classes analyzed in SNAP software. Excel for graph plot
Scene
241.04 m
5. METHODOLOGY 5.1 Pre-Processing of Air-Borne and Space-Borne Sar Data Data acquired by the air-borne and space-borne SAR sensors contain uncertainties due to variations in altitude and velocity of the sensor platform. The datasets need to be properly calibrated/preprocessed to use it for desired applications. Preprocessing of the SAR data includes conversion from slant range to ground range, radiometric calibration, speckle suppression and SAR image geocoding. Since SAR image has complex geometry and backscatter value is in complex and imaginary part, the data requires preprocessing in order to be understood in its entirety.
Pre-processing and backscatter image generation from RADARSAT-2 datasets is carried out following the steps mentioned below:
Backscatter for a target area at a particular wavelength will vary for a variety of conditions, such as the physical size of the scatterers in the target area, the target's electrical properties and the moisture content, with wetter objects appearing bright, and drier targets appearing dark. The exception to this is a smooth body of water, which will act as a flat surface and reflect incoming pulses away from the sensor. These bodies will appear dark. The wavelength and polarization of the SAR pulses, and the observation angles will also affect backscatter
Figure 7 Methodology to convert SLC Image Pixel value to dB values
We refer to three types of reflection, which represent the three extreme ends of the way in which energy is reflected from a target: specular reflection, diffuse reflection and corner reflection.
Specular reflection When a surface is smooth, we get specular or mirrorlike reflection where all of the energy is directed away from the surface in a single direction
Figure 6 Methodology to analyse through Polarimetric Decomposition Figure 8 specular reflection
The overall study divided into two parts namely part1 and part-2. In phase-1, backscatter image analysis has been carried out and in part-2, Polarimetric analysis has been carried out.
6. ANALYSIS PART 1 In remote sensing, we are most interested in measuring the radiation reflected from targets. SAR images represent an estimate of the radar backscatter for an area on the ground.
Figure 9 specular reflection on smooth surface
Diffuse reflection occurs when the surface is Darker areas in the image represent low backscatter, while brighter areas represent higher backscatter. Bright features mean that a large fraction of the radar energy was reflected back to the radar, while dark features imply that very little energy was reflected.
rough and the energy is reflected almost uniformly in all directions. Most earth surface features lie somewhere between perfectly specular or perfectly diffuse reflectors Figure 10 Diffuse reflection
LAKES Lakes show up very Dark in a radar image. When a surface is smooth, we get specular reflection where all of the energy is directed away from the surface in a single direction all reflation goes in other direction small amount of energy is returned that’s why value comes like -20 dB to -30 dB Figure 13 Diffuse reflection
Corner reflectors The orientation of the surfaces at right angles causes most of the radar energy to be reflected directly back to the antenna due to the double bounce reflection. Corner reflectors in urban buildings and streets, bridges, other man- Figure 11 Corner reflectors made structures
Feature Class
AOI 1
Lake Backscatter value
-23.2
AOI 2
AOI 3
AOI 4
AOI 5
-23.5
-24.2
-25.1
-23.4
Five AOI’s were taken from three lake and the values have a mean of -23.92 dB there is no object on surface so only reflection happen.
7 RESULTS URBAN In Urban areas, buildings appear very bright in a radar image of a town because the building sides act as corner reflectors.
Figure 12 Backsacatter values for Urban Features Feature Class Urban Backscatter value
Figure 14 Backscatter Values for Lake
AOI 1
AOI 2
AOI 3
AOI 4
AOI 5
3.362
1.795
1.671
3.157
2.783
FOREST Most surfaces are not smooth, and reflect incoming EM radiation in a variety of directions. These are called diffused reflectors or scatterers. When waves interact with only the foliage, it is called diffused reflection. Feature Class Forest
AOI 1
AOI 2
AOI 3
AOI 4
AOI 5
Backscatter value
-13.04
-15.71
-11.68
-9.011
-12.15
Feature Class Cultivated land Backscatter value
AOI 1
AOI 2
AOI 3
AOI 4
AOI 5
-5.444
-8.004
-8.570
-5.77
-6.82
8. Conclusion
ROAD EM Radiation from very smooth surfaces follows Snell’s laws of reflection and hence surface scattering occurs in case of roads. Therefore, the backscatter values are very low. (Below -20dB). Feature Class Road Backscatter value
AOI 1 3.362
AOI 2 1.795
AOI 3 1.671
AOI 4 3.157
AOI 5 2.783
To interpret land features in SAR Images, five AOI’s for each class have been taken in SNAP and their average found. Then, each of these values has been analyzed in HH Polarization Table 1 Land use class features with average dB
S.R. No. 1 2 3 4 5
Land features
Average dB value
Lake Forest Road Cultivated land Urban
-23.9217 -12.3184 -15.8512 -6.92219 2.554082
Figure Figure1615Backscatter BackscatterValues Valuesfor forForest Road(Dense Vegetation)
CULTIVATED LAND Cultivated land contains growing/grown crops which reflects in multiple direction and hence has low value of backscattering as compared to Urban Features.
Figure 17 Backscatter Values for Cultivated Land
Figure 18 COMBINE LINE PLOTE IN HH POLARIZATION
By analyzing this it is clear that Urban features contributes very high backscatter intensity (2.554082 dB). Waterbody contributes lower backscatter intensity (23.9217 dB). All the other features (viz. Agricultural Lands (~6.92219 dB), Forests (-12.3184dB), Roads (-15.8512 dB), etc.) contribute to higher backscatter intensity than water but lower than the Urban targets. The amount of backscatter in energy is in negative because the amount of reflected energy from the target is lower compare to incident energy.
In the previous monochromatic image of Vancouver, the tones of grey describe the scattering intensity of various land target features. •
•
Figure 19 AVERAGE BACKSCATTER VALUE IN HH POLARIZATION
•
Targets appearing in white colour have strong scattering intensity like urban area, ship, bridge. Targets appearing in shades of grey have medium scattering intensity and can be attributed to crops, forests etc. Targets appearing in black have low scattering intensity and are smooth surface such as road, water, etc.
9. ANALYSIS PART 2 9.1 Span Image: Also Known as Power Index, it can be expressed as the sum of square of intensity of four polarization channels, namely, HH, HV, VH and VV. The values of intensity for a span image vary between the features of water and urban. Properties of the surface with which the energy interacts play a major role in determining the intensity values. Thus, features with high backscatter value appear bright and possess high intensity and those with low backscatter value possess low intensity.
Figure 20 Span (Total Power) Index Image
Figure 22 Colour Slider for Span Image Histogram
Figure 21 SPAN image of the study area
SPAN image has been classified with 8 colour gradient which helps us to classify targets better. •
•
•
Lake
Red colour shows features with strongest scattering intensity (-5.54 db to +16.683) in this urban, ship,bridge . Shades of green shows features with medium scattering intensity (-11.74 to – 5.54 intensity dB). White and shades of blue shows features with lowest scattering intensity.
The slider has a maximum value of 16.683 and minimum value of -23.191 for the given Span image of Vancouver. Features appearing in yellow colour have the highest count. Various AOIs for different targets have been taken on the SPAN image and their backscatter values have been noted. Scatter plot shows that ships (9.6327 intensity dB) have strongest scattering intensity followed by urban areas (7.1065), bridges (4.63), Dense vegetation (-3.52), Moderate vegetation (2.198) , fallow land (-8.023), Water with sediment (14.2615) and waterbody is the least with scattering intensity of -20.44.
Figure 26 Lake in Span dB Figure 23 Lake in Google Earth Image Image
The span dB intensity of water is -20.44 db. Water has the lowest of backscatter value resulting in a negative mean backscatter value. AOI’s taken represent calm water and returns reflectance value lower than the incident reflectance.
Bridge
Dense Urban The span intensity of Urban is 7.1065 db. Urban targets have the highest backscatter value resulting in a positive backscatter value.
Figure 27 Bridge in google earth image
Figure 24 Urban In SPAN dB
Figure 28 Bridge in Span dB image
Figure 25 Vancouver Urban Areas From Google Earth Image
The span intensity of bridges is 4.6308 db. Bridges have a high backscatter resulting in a positive backscatter value. This coincides with the image classification of urban land-use class where all urban features had maximum backscatter value.
Ship The span intensity of ships is 9.632 db. Ships have a high backscatter value resulting in a positive backscatter value. Figure 30 ships in Span dB image Ships have high backscatter value and can easily be spotted in the calm water which has least reflective backscatter Figure 29 ship in google earth value.
Water has the lowest of backscatter value resulting in a negative mean backscatter value. Water with sediment appear in shades of blue and cyan in SPAN image. The span intensity of water with sediment is 14.261 dB. Water of the study area has sediment and has higher reflectance than still water
Fallow land
image
Road Figure 36 Fallow land in span dB
Figure 31 Road in Span dB
Figure 32 Road in google earth
The mean span intensity of roads is -10.2565 dB. Roads have a homogeneous smooth texture and have surface backscatter reflection. They will have negative backscatter reflection.
Water with sediment
Figure 33 Fallow land in google earth
Fallow land has lower scattering values as compared to Urban and thus appear in Green in SPAN Image. The span intensity of fallow land is -8.023dB. Fallow land has a homogeneous smooth texture and hence has diffused backscatter reflection. They will have negative backscatter reflection. Target
Figure 34 Water with Sediment in Span dB
Dense Urban Ship Bridge Road Fallow Land Water with sediment Lake
Figure 35 Water with Sediment in Google Earth Image
Backscatter Response 7.106515263 9.632745854 4.630873704 -10.25650121 -8.023462653 -14.26156586 -20.44259481
TARGET
Dense Urban
Figure 37 Backscatter Responses of individual targets
The span gives the overall behavior the surface to backscatter and hence rough geometries would give high back scatter span values while smooth surfaces such as Road, Lake water would produce low span values in DB.
9.2. Volume Scattering Index:
MEAN VOLUME SCATTERING -3.46385
Forest
-5.32982
Terrain slop Ship
-5.35174
Sparse urban
-6.88665
Bridge Crops
-8.02494 -9.10235
Water with sediments
-11.7683
Water
-12.8584
-5.61385
INFERENCES
High volume scattering, dense area High volume scattering, dense area High volume scattering High volume scattering Less dense area, medium or low height buildings Rough Surface Medium volume scattering Low volume scattering because of smooth surface Low volume scattering because of smooth surface
Dense Urban The volume scattering index intensity of Dense urban area is -3.46385 db.
Figure 38 Classified Volume scattering index image
Volume Scattering Index (VSI) is the measure of depolarisation of a polarised incident radar signal. Higher values of VSI occur when cross-polarization backscatter component is greater than the copolarization backscatter component in the depolarised signal. This depolarising mechanism is also known as multiple-path scattering. The above image shows volume scattering index of various targets in the study area. Features appearing in shades of orange and red have high volume scattering while features appearing in white and shades of blue have low volume scattering. High volume scattering is the result of roughness of the surface, heterogeneity of the area, multiple scattering etc.
This could be due to the presence of dense urban clusters in downtown area of Vancouver which show higher values of volume scattering and consequently appear brighter in the image because crosspolarization is greater than the co-polarization backscatter more rough surface.
Sparse urban The volume scattering index intensity of Sparse urban area is -6.8867 db. The Sparse urban areas have smaller houses producing a diffused effect and hence have a comparatively lower scattering value because cross-polarization is greater than the co-polarization backscatter more rough surface.
Water The volume scattering index intensity of shallow water is -12.8584 db. The points of interest were taken from still water lakes where the water was calm and hence has diffused volume scattering acquiring backscatter
intensity in the lowest end of the spectrum and more smooth surface co-polarization backscatter is greater than the cross-polarization.
Water with sediments The volume scattering index intensity of water with sediment is -11.7683 db. The points of interest were taken along the coast. The reflectance intensity of water is the lowest but in water with sediment gives us a brighter backscatter and higher mean value of 11.7683 dB due to increased roughness because copolarization backscatter is greater than the crosspolarization
Forest The volume scattering index intensity of forest is 5.32982db. Because of density of vegetation foliage over a larger area, it shows higher volume scattering more rough surface cross-polarization is greater than the co-polarization backscatter.
Observations from Volume Scattering Index Figure 39 LINE PLOT OF VOLUME SCATTERING INDEX
target will appear red on the image, i.e. it will have double bounce. If the feature is parallel or at a 45° angle, it will appear green or indicate volume scattering.
Scattering Mechanisms Double Bounce Surfaces inclined towards the radar will have a stronger backscatter than surfaces which slope away from the radar and will tend to appear brighter in a radar image. Some areas not illuminated by the radar, like the back slope of mountains, are in shadow, and will appear dark. When city streets or buildings are lined up in such a way that the incoming radar pulses are able to bounce off the streets and then bounce again off the buildings (called a double-bounce) and directly back towards the radar they appear very bright (white) in radar images. Roads and freeways are flat surfaces and so appear dark. Buildings which do not line up so that the radar pulses are reflected straight back will appear light grey, like very rough surfaces.
Double-bounce Scattering:
By the interpretation of line plot, it is clearly visible that high backscatter value represents urban, forest, ship and terrain slope. The lower back scatter values are observed in waterbodies and water with sediment. Whereas, Bridges, Crops, etc. show medium backscatter values.
9.3. Freeman Durden Decomposition: The Freeman Durden decomposition method uses the T3 coherency matrix where T11 is odd bounce or surface scattering. T22 is even bounce or double dihedral scattering and T33 is diffuse bounce or volume scattering. The odd, even and diffuse bounce are seen in the red, green and blue bands respectively. For urban targets of any tall standing features; the targets that are in a plane perpendicular to the radar the
Figure 40 Freeman Durden Decomposition - Double Bounce
Volume Scattering: Figure 44 Freeman Durden Decomposition - Volume Scattering
Above image shows the Freeman Decomposition – Double Bounce Band.
Durden
It highlights the features on the study area which have maximum double bounce scattering values. Such features have been shown in red box. As we can see, urban areas with medium or low height buildings and ships majorly have been highlighted in this image. Other features appear dark. It gives the Double-bounce backscatter intensity. Generally, man-made features contribute the higher intensity (Above -5 dB)
Figure 41 Overall Trends in Double Bounce across all feature classes
Figure above shows the Freeman Durden Decomposition – Volume Scattering Figure 43 Feature Classes with high Double Bounce Intensity
Band. Figure 42 Feature Classes with High Volume Scattering Intensity
It highlights features which gives maximum values in Volume Scattering. Features heterogenous in nature or dense areas give high volume scattering like urban sprawl, forests etc. All other features having high double bounce or surface scattering values have been suppressed. It gives the Moderate backscatter intensity. (-20dB to -10dB). In some case it contributes high backscatter intensity say below -1dB in case of dense urban.
Above figure shows Freeman Durden Decomposition – Surface scattering. It highlights features giving maximum values for surface scattering while suppressing other features.
Figure 45 Overall trends in Volume Scattering across all feature classes
Surface Scattering:
Flat surfaces, waterbodies, smooth surfaces give high values for surface scattering while rough water gives low surface scattering values. It gives the lower backscatter intensity. (Below -15 dB)
Figure 47 Overall Trends in Surface Scattering across all feature classes
Figure 48 Feature Classes with High Surface Scattering
Figure 46 Freeman Durden Decomposition - Surface Scattering
Figure 49 Comparing value of features in all three type of scattering in Freeman Durden Decomposite
Figure 50 Classified Volume scattering index image
Freeman Durden Decomposition – RGB assigns colours to the features in the study area according to their backscatter values. Band composition given is – surface in blue: volume scattering in green: Double bounce in red that is because double bounce features reflect highly in red band while, volume scattering features in green and surface scattering features reflect better in blue. Thus, high rise urban, ships appear in red, low rise urban, forests, crops appear in green and water bodies, roads, wet farms, snow appears in blue.
10. Conclusion The main focus of the present study was to analyze the backscatter values of various identified feature classes from the C-Band Radarsat 2 imagery of Vancouver City. It is observed that water shows minimum backscatter values while the urban land features show maximum backscatter value intensity. The scattering plots and the polarization response of various land feature class helps us understand and justify the various identified classes. After Image interpretation, one can understand the various applications of SAR Imagery such as being used for crop monitoring, ship monitoring, etc. It has innumerable other applications due to its all-weather capability. It can be concluded that SAR images give us more information due to better interacting properties.
11. References [1] Natural Resources Canada, “Fundamental of Remote sensing,” NARCAN, Canada [2] RADARSAT-2 Product Description, Issue 1/13: March 21, 2016. Figure 51 Histograms Freeman-Durden decomposition
[3]
https://earth.esa.int/documents/10174/2700124/sar_la nd_apps_1_theory.pdf [4] https://www.microimages.com/documentation/Tutorials/rad ar.pdf
[5] http://nile.riverawarenesskit.org/English/NRAK/EO/ html/rsbch12.html [6] https://arset.gsfc.nasa.gov/sites/default/files/water/Br azil_2017/Day1/S1P3.pdf