DEM Construction using DINSAR

ABSTRACT: A digital elevation model (DEM) is a 3D visualization of a terrain surface. It can be used in various analytical studies such as topographic feature extraction, hydrology, geomorphology and landslides analysis etc. Uttrakhand region is affected with landslides, earthquake and flash flood phenomenon. Hence this study was focused on DEM generation using Differential SAR Interferometry (DINSAR) on ALOS PALSAR dataset. Two Pass DINSAR technique involves one interferometric pair in addition with an external DEM. The external DEM was used as a reference to reduce topographic errors. The data processing steps were image co-registration, interferogram generation, interferogram flattening (Differential Interferogram), interferogram filtering, coherence map, phase unwrapping, orbital refinement and re-flattening and DEM generation. Interferogram fringes observed in forest areas were due to temporal decorrelation and the fringes in mountain regions were obtained due to topography changes (may be due to landslides in rainy season). The range of elevation in generated DEM were 132m to 2823m and Root Mean Square Error (RMSE) error was 36.765159m. The generated DEM was compared with ASTER DEM and variation in height was analyzed. Atmospheric effects were not removed due to geometrical and temporal decorrelation which affect the accuracy.


Introduction
In Uttrakhand region, landslides, floods and earthquake are very common phenomenon which leads to loss of lives as well as assets.To analyze these problems, Digital elevation model (DEM) is required.DEM is a 3-D representation of earth surface.It gives elevation value at every (x, y, z) coordinate on the surface.It can be constructed in many ways through stereo pairs (Toutin and Chang 2001), GPS ground data, contour maps (Carrara et al 1997), Lidar data (Axelsson 2000) and SAR Interferometry (Rosen et al 2000).Collection of ground data (DGPS measurements for DEM generation) and contour map generation required much intensive field work and this task is time consuming.For quick detection and prevention from disaster, Synthetic Aperture RADAR (SAR) Interferometry is being used globally.
SAR is an active microwave imaging system.It has its own source of radiation which can be used throughout day and night.SAR sensor can penetrate the cloud cover and can reach the depth in the subsurface.(Richards 2007) The phase difference is calculated by subtracting the  1 from  2 as shown in equation ( 2.3) Where λ is the wavelength of sensor.
From the cosine law: Where B is the Baseline, θ is the look angle and α is the angle between the parallel baseline  ℎ and original baseline B.
Neglecting ΔR 2 (very small), reshuffle the equation:  Phase can be calculated using the equation (5.2): The phase difference in interferogram image is due to many sources explained below: Where ∆  = flat earth phase, ∆  = topographic phase, ∆  = phase due to atmospheric disturbances, ∆  =phase due to decorrelation noise, ∆  = phase due to ground displacement between two SAR images.

Interferogram Flattening (Differential Interferogram):
The phase calculated in interferogram is affected due to earth curvature.The process of its removal is called Interferogram flattening.This provides us with fringes which are related to change in displacement (if any), elevation, noise and atmospheric effects.The flat phase can be estimated as shown below: Where   = look angle to each point in the image assuming zero local height

Interferogram Filtering and Coherence Generation:
Filtering reduces the noise and smoothens the interferogram.
Coherence map is generated from the filtered interferogram.It measures the degree of similarity between two pixels.The mathematical formula is given below: (Fletcher, K. 2007) Where N is the no of pixels, M1 is master image, S1 is the slave image and S1 * is the complex conjugate of slave image.
The significance of coherence map is to determine the quality of interferometric pair.Low coherence means the data is not suitable for DEM generation.

Phase Unwrapping:
The range of interferogram phases is (-π, π) only.This process is used to add the correct multiple of phases which resolves the 2π ambiguity.(Reigber and Moreira 1997) Where   is the wrapped phase,   is unwrapped phase and n is a positive integer.

Orbital refinement and re-flattening:
This step refines the orbit values, removes the phase ramps and calculates the phase offset values.This is compulsory step for DEM generation.Ground control point file is required to execute this step.

DEM Generation:
The calibrated and unwrapped phase value is converted into height value and geocoded into map projection.The height value is calculated using normal baseline method expressed in equation (5.7).

RESULT AND DISCUSSION
Baseline: The normal baseline of data-pair is less than the critical baseline as shown in Table 1.It means data pair is suitable for DEM generation.

CONCLUSION
The goal of this study was to generate the DEM using Differential SAR Interferometry.Two dual polarized ALOS Palsar dataset were used along with 90m resolution external DEM (SRTM).From the results, RMSE errors were found very less in urban areas due to high correlation whereas more errors were obtained in mountainous regions (may be due to landslides in rainy season) and forest regions due to temporal decorrelation.These errors could be further reduced by using 30m resolution external DEM.
Atmospheric effects were not removed with this technology.Further improvements can be done using PSINSAR and SqueeSAR technologies.
SAR Interferometry allows measurement of radiation's travel path accurately because of coherent nature of microwaves.Measurement of phase difference helps in generation of DEM. ( Semwal 2013) This study focusses on DEM generation using Differential SAR Interferometry (DINSAR) technique.The main goal was to analyze the accuracy of generated DEM with ASTER DEM.The description of technique is mentioned below.2. DINSAR Technique Two Pass Differential Interferometry is an advanced version of SAR interferometry in which two interferograms are generated.One from SLC pair and another interferogram corresponding to surface topography based on external DEM.Finally the difference of phase information of two interferograms are used to generate Differential Interferogram as shown in Fig 1.

Fig 1 :
Fig 1: Two-Pass DINSAR Technique ( Balaji 2011) 3. Mathematical Background Consider S1 and S2 are the positions of image radar antenna separated with baseline B as shown in Fig 2.  ℎ and   are the parallel and perpendicular baseline respectively. and  +  are the ranges from the ground.(Chatterjee, R.) Fig 3: Study area (LISS III imagery in standard FCC)

Fig 4 :
Fig 4: Flow diagram of Methodology Baseline Estimation: It provides information about baseline values and orbital parameters of interferometric pair.This information is not used for DEM generation but provides the feasibility of data.Co-registration:Co-registration is defined as a process of superimposing two or more SAR images in the slant range geometry.External DEM is used for this purpose.(Li and Bethel 2008) Interferogram Generation: Each pixel of SLC image has amplitude as well as phase information.The interferogram is generated by Hermition product of each pixel between two co-registered images.The formula is shown below in equation (5.1):   =   *   * Fig 4: (a) Interferogram (b) Flattened Interferogram (c) filtered Interferogram (d) coherence map (e) Wrapped Phase DEM Generation: Generated DEM was of 25m resoution having elevation values between 132m to 2823m as shown in Fig 5.The RMSE error was calculated using three degree polynomial function was 36.765159m.The newly constructed DEM was than compared with ASTER DEM having elevation values between 193m to 2731m as shown in Fig 6 and it was found to have very less variation in settlement areas and high variation in mountaneous and forest areas.

Fig
Fig 5: Digital Elevation Model

Table 1 :
Baseline Data Pair Fringes absorbed in the image are the phase changes due to topography, atmospheric effects, noise and displacement (if any) shown in Fig 4(a).Phase difference due to curvature of earth was removed.Phase discontinuity can be seen as contour lines as shown in Fig 4(b).