Lineament Extraction From Quad-PolSAR Image In The Area Of Nanhu Gobi Desert In Hami, Xinjiang

Many kinds of shallow cover areas (such as the Gobi desert, vegetation cover, loessial overburden area, glaciation, etc.) widely distribute in the land of China, which may reserve rich mineral resources. However, due to mulching or semi-mulching layer, real geological and ore information transformed into "indirect information", "weak information", "composite information", and "incomplete information". Synthetic Aperture Radar (SAR) can penetrate the shallow cover and detect the information of lithology and geological structure under the cover. With the rapid development of SAR technology, Quad-PolSAR provides the promising new possibilities of geological exploration. However, previous geological studies are focused on lithological identification and classification by using quad-polarization SAR and lack of recognition and extraction of geological structures under the coverage area. Therefore, there is a great need for further research on detecting geological structures with polarization SAR, which tends to expand the potential applications of Quad-PolSAR.Spaceborne Quad-PolSAR has the characteristics of side-looking and the ability of penetrating, which have been used to mapping both subsurface lineaments and surface lineaments in the zone of Nanhu Gobi desert of Hami, Xinjiang Province. C-band RADARSAT-2 and L-band ALOS-PALSAR data are utilized to semi-automatically extract the lineaments and access. This paper is focused on preprocessing of topographic effect in SAR imaging, lineaments identification and lineaments extraction.Firstly, Geometry distortion of SAR image induced by side-looking imaging, radiometric distortion and effect of polarization azimuth angle shift have been preprocessed. And their influence to position, orientation and scattering signal of lineaments also have been analyzed.The methods of simulation DEM imaging and compensation of polarization orientation angle were taken to achieve the correction of geometric distortion and radiation distortion, particularly roll-invariant polarization parameters are introduced to maximally inhibit the effect of polarization orientation angle shift on scattering of lineaments. Secondly, different kinds of physical scattering model of lineaments are built according to the frame of polarization signal and polarimetric decomposition; Edge persevering index and equal number of looks are used to assess these polarization parameters for linear feature extraction. Finally, in order to achieve the main direction of local linear feature on SAR image, histogram statistic of gradient based on circle neighborhood (HOG, Histogram of Oriented Gradient) and particle filter tracking algorithm are applied in this paper. Then the analysis and assessment of volume scattering and roll-invariant polarization parameters are presented to show their application prospect in lineaments mapping and structural geology.The main conclusions of the article are listed as follows:(1) Lineament identification and topographic distortion correction. The results show that the preprocessing process of geometric distortion must be properly treated and corrected by using stimulated DEM imaging for SAR. However, the radiation distortion caused by the terrain is beneficial for lineaments extraction, which highlights regional topographic features. In addition, the polarization angle shift induced by varied orientation of lineaments can be corrected by compensation (use circular polarized algorithm to estimate the polarization orientation angle shift, then rotate scattering matrix with that angle), and the compensated peak mainly distributed in texture-rich region that cased by topographic relief. Roll-invariant parameters have good performance in restrain the effect of that shift.(2) Analysis of polarization signals and scattering mechanism for lineaments. Synclines, ridgelines, valleys and covered faults are the main types of lineaments in the area of Gobi desert. Synclines and ridgelines are characterized as strong backward scattering in single or odd mode on the SAR image. Due to the large local incidence angle, small surface roughness, valleys are shown in dark linear features on SAR image and produce the bragg backscattering. The lineaments in the coverage area also has a weak volume backscattering. After feature selection based on the analysis of physical scattering mechanism, edge persevering and noise level, the odd scattering component, volume scattering component, components T11 that corresponds to the maximum eigenvalue based on eigenvalue decomposition and T11 (roll-invariant parameter)of coherent matrix are the most optimum parameter for lineaments extraction.(3) Lineament extraction and pattern analysis of geological structures. Line tracking algorithm is applied to detect different types of lineaments and achieve a brilliant effect on target searching. Circular neighborhood of HOG statistics is isotropic, which is good at directional prediction by weighted gradient statistics under the interference of noise. However, certain directional predictive error and searching interruption may occur when multiple edges intersect. Neighborhood size and searching step are two most important factors that affect lineaments extraction. At last, the extracted lineaments are utilized to analyze the pattern of regional geological structures and update the previous geological survey. It can be conclude that the pattern of geological structures is consistent with previous survey and get more detailed lineaments in the covered area.