The Detection And Reconstruction Of Urban Objects Using Polarimetric Sar Image

Synthetic aperture radar (SAR) imagery is one of most advanced technologies during recent two decades in microwave remote sensing, especially in the urban remote sensing. As the development of the polarimetric SAR and high resolution SAR, the SAR technology has been used in more aspects of the urban remote sensing, such as the monitor of the city development and change detection. As the resolution reaches a meter order, reconstruction of the 3D stereo objects on the terrain surface from all-weather SAR image becomes more feasible. Due to some technical limitations, conventional radar measurements are mostly restrictive to well classifying, inverting and reconstructing the stereo objects. It is remained for further study via fully polarimetric SAR imagery with high resolution.Since pioneer experiments of SIR-C in 1994, some airborne and satellite-borne polarimetric SAR have been developed. The satellite-borne ALOS phase array polarimetric SAR at L band has been launched in January 2006. Canadian Radarsat-2 polarimetric SAR at C band will be in orbit in early 2007. It is expected how to effectively implement broad applications of polarimetric SAR images has become an extensive topic. The PI-SAR is a Japanese airborne polarimetric and interferometric SAR with two independent radar systems (L band-1.27 GHz, 3m resolution, and X band-9.55GHz, 1.5m resolution). Interferometric data at X band and only polarimetric data at L band can be obtained. Fully polarimetric data from the square loop flights of PI-SAR over Sendai city, Japan in 2004 becomes available. Multi-directional flights can see the 3D objects in different angles and lead to reconstruction of those stereo objects.In this paper, the radar image is studied and an conclusion is made that the difference between the real location of an object and its location on the radar image is relevant to the height of the object and this relationship can be used to calculate the geometrical index of the object and therefore rebuild it. Employing the images of coherency matrix components derived from polarimetric PI-SAR images of two converse flights at X band, which are mostly characterized by double bounce scattering between the building sides and ground surface, a feasible approach to reconstruct the 3D stereo objects, such as a TV tower and campus buildings is developed. Reconstructions of the building targets are validated by the real ground truth.Full-polarimetric SAR image can provide much more information containing the scattering characteristics of an object that can be used to classify the image and also detect objects from polarimetric radar images. By researching different types of objects a it can be concluded that different components of radar data can represent different kind of scattering model. Based on this, different components of polarimetric radar data should be chosen implying to detection of different kinds of objects.CFAR is often used to detect radar objects automatically. Using the estimation of the radar noise the adaptive threshold can be calculated to determine the pixel belongs to an object or not. Based on the former CFAR, a new adaptive multi-scale CFAR is employed that is more appropriate for the objects detection in urban areas where the environment is much more complicated with many kinds of objects of various shapes and sizes. The new employed CFAR can be used to detect the objects of various sizes that are often seen in urban areas, such as buildings and automobile objects. Besides, the study of the background regions is adapted to estimate the radar noise power more accurately. Then the objects detected can be matched automatically on the basis of the size of the objects in the radar images of the different flights. Based on the different sizes of the objects in the urban area, two typical kinds of objects (automobiles and buildings, which are small and comparatively big) can be classified and rebuild respectively and therefore the urban environment can be reconstruct when ignoring other kinds of objects.