Title : Automatic method used for geometric correction for SAR Abstract Radar imagery has become one

of the most important data sources and efficient tools for terrain analysisand natural resource surveys since 1960s. With the development of technology in the field of radar remote sensing, new generation of radar sensors, i.e., Synthetic Aperture Radar (SAR) was born. Unique specifications of radar systems and images versus optical ones led to a whole new series of applications for radar imageries all over the world. However, the level of achievable accuracy from radar imageries is still a problem for their applications. Multiplicative noise such as speckle which is unavoidable part of coherent radar images, degrade radiometric quality and interpretability. Moreover, geometric distortions such as foreshortening, layover, shadow and other problems related to special imaging geometry of radar systems, decrease reliability of radar imageries. Thus, radiometric and geometric corrections and calibrations must be applied to the radar images before using them. Introduction Radar remote sensing, like optical remote sensing, is used to produce the image of Earths surface. A radar image is a record of the interaction of energy and objects at the Earths surface. Its appearance is dependent on variables such as geometric shape, surface roughness and moisture content of the target object, as well as the sensor-target geometry and the transmission direction (look direction) of the radar sensor. There are significant differences, however, between how a radar image is formed and what is represented in that image compared to optical remote sensing imagery [3]. In compare to optical

remote sensing, radar imaging has some advantages. First, as an active system, it is a day/night data acquisition system. Second, considering the behavior of electromagnetic waves in the range of RADAR wavelength, it can be seen that atmospheric characteristics such as cloud, light rain, haze, and smoke has little effect on the capability of RADAR data acquisition system. This makes RADAR as an allweather remote sensing system. Last but not least, as the RADAR signals partially penetrate into soil and vegetation canopy, in addition to surface information, it can provide subsurface information too. The returned signal (backscatter) from ground objects (targets) is primarily influenced by the characteristics of the radar signal, the geometry of the radar relative to the Earths surface, the local geometry between the radar signal and its target, and the characteristics of the target. Content Radar systems are side-looking distance measuring systems, thus key geometric parameters are the incident angle, local incident angle and look direction. The sidelooking geometry of radar results in several geometric distortions, such as slant range scale distortions and relief distortions. Geometric corrections include slant to ground range, registration, and local incident angle corrections (if topographic information is available). Generally speaking, geometric correction algorithms are classified into three methods: Slant to ground method Polynomial method (best fit approximations) Radargrammetric method (known sensor geometry) Ground Control Points (GCPs) are used to establish and/or refine the transformation.

Slant/Ground Range Conversion SAR data are acquired in slant range. Slant to ground range conversion is used to project the acquired image to the ground system. To do this, one needs to know (or assume) imaging geometry, platform altitude, range delay and terrain elevation. Resampling is used to give uniform pixel spacing (in ground range) across the image swath. Slant to ground range conversion can be done during signal processing or during image processing. Generally, it is applied after radiometric correction. Approaches and algorithms used are a function of analysis objectives. RADARSAT ground range products assume a sea level ellipsoid earth model with zero relief. Image Registration Polynomial Transforms Polynomial transform uses a best-fit model. First order polynomial is a shift-rotation of the image, whereas the third order polynomial is a complex warping of the image. Second order polynomials are used for images requiring nonlinear warping. Third and higher order polynomials create a more complex image transformation. Higher order transforms require a greater number of ground control points (GCPs) in order to produce the transform model. High order does not guarantee higher accuracy. Higher order usually ties the image down at the GCPs, but can increase errors between the GCPs. Radargrammetric Method Geocoding is the geometric correction of image data to a map projection. Traditional method of geocoding is the polynomial transform. This method does neither model the viewing geometry nor use elevation data to correct for topography. The most accurate geocoding method is the radargrammetric

method. The radargrammetric consists of three steps as following:

process

Ephemeris modeling and refinement (if GCPs are provided) Sparse mapping grid generation Output formation corrections) (including terrain

Radargrammetric method uses analytical formulation of the distortions during image formation. Therefore, the geometric correction is done using the platform (ephemeris and ancillary data), sensor (integration time, pulse length, depression angle), and DEM information. Output of radargrammetry is an Ortho-image corrected for all distortions, including relief. The planimetric accuracy of the final orthoimage is dependent on the accuracy of GCPs and the DEM.

The advantages of radargrammetric method are as following: Unified projection system. Direct image to terrain correction. Only one resampling of an image (slant range to map projection is directly done, no intermediate conversion to ground is required). Homogeneity in the ortho - image generation. Use of a DEM or a mean altitude. Better integration with GIS or digital maps. Comprehension and control of the full geometric process and of the resulting errors.
Conclusion Geometric corrections in radar imageries are different than optical ones as the geometry of the imageries are different. The radargrammetric method has a better performance in compare to the other two

methods. The reason seems to be obvious as radargrammetry considers geometry of imaging, uses both orbital parameters of the sensor, and DEM of the region.