Keywords: Polarization Signature, Textural Feature, PI-SAR
Abstract For high resolution SAR images, the influences of the factors that can be ignored for usual SAR images, such as the shadows of buildings and trees, have to consider. In this paper, the influences of the factors particular to high resolution images for polarization signatures and textural features, and the changes by the differences of wavelength are analyzed using PI-SAR images.
1. Introduction
Recently, resolution of remote sensing image data has been higher, and the tendency has been seen not only for visible sensor images but also for Synthetic Aperture Radar (SAR) images. For the observations of earth using SAR, to observe the detailed physical properties for the surfaces has been possible by polarimetry technique, and various products have been reported [1],[2]. Analysis of polarization signatures, one of general methods for the analysis of polarimetric SAR images, has been expressed in many papers and shown its availability [3]. Textural analysis has also been carried out for polarimetric SAR [4],[5]. However, it is scarcely discussed that changes polarization signatures and textural features caused by improvement of the resolution.
In this study, the influences carried by the shadows of buildings and trees that have never been considered in usual resolution polarimetric SAR (SIR-C and so on) are analyzed using airborne PI-SAR (Polarimetric and Interferometric SAR) high resolution images. In general resolution SAR images, the shadow effect should be considered for the analysis of mountain areas. For these images, the shadows of buildings and trees had been ignored because their sizes were smaller than limits of resolutions. However, unusual surface patterns are formed clearly due to the fine shadow included the analysis areas using higher resolution images. The influences of back-scattering coefficients for the patterns are compared using polarization signatures and the influences of image structures are analyzed using textural features. For PI-SAR data, the resolution varies according to the frequency band (X-band or L-band), so that the resolutions are transformed the same value and the differences among wavelength are also discussed.
2. Pi-Sar Data
PI-SAR is one of airborne SAR developed by CRL and NASDA for the purpose of observing earth environment and watching disasters [6],[7]. Its observation frequency band are X-band (9.55 [GHz]) and L-band (1.27 [GHz]), and the resolutions vary according to the wavelength; 1.5 [m] and 3.0 [m] respectively. In distributed PI-SAR data products, there are SSC (Single-look Slantrange Complex), MGA (Multi-look Groundrange Complex), and MGP (Multi-look Groundrange Polarimetric) formats. In this study, MGP format data was selected because this format data has full-polarimetric information. For each pixel, the polarization information expressed by 10 byte data.
As the object areas, vegetation area and residential area were mainly extracted. In these areas, though shadows of trees and buildings appear scarcely, the changes of pattern influenced by the shadows should be considered for high resolution sensor like PI-SAR. The overview of PI-SAR data used in this study is shown in Table 1, and the sizes of extracted areas are shown in Table 2 respectively. Fig.1 shows the positions of these areas in used scene.
Table 1: Overview of used PI-SAR data
| Subject |
X-band data |
L-band data |
| Frequency [GHz] |
9.55 |
1.27 |
| Azimuth resolution [m] |
1.5 / 3.0 |
3.0 |
| Range resolution [m] |
1.5 / 3.0 |
3.0 |
| Data format |
Multi-look Grouondrange Polarimetric |
| Observation site |
Tsukuba, Japan |
| Acquisition date |
1997. 9. 30 |
Table 2: Sizes of extracted areas (Unit: pixels)
| Name of area |
Azimuth orientation |
Range orientation |
| Paddy field |
140 |
140 |
| Vegetation area |
160 |
160 |
| Residential area |
170 |
130 |
Fig.1: Extracted areas in Tsukuba site (X-band HH polarization image)
© Communication Research Laboratory 1999.
