Dual-band SAR Backscatter And Coherence Analysis And Applications On Land Use/Cover Classification

Because of it’s all-weather, all-time detection capabilities, active synthetic apertureradar (SAR) has great potential for applications in the regions of cloudy and rainy weather.With the enrichment of different polarization and different band SAR data, it is necessaryto compare and analysis characteristics and application potential with different bands anddifferent polarization SAR data, in order to synthetically use the advantage of differentbands and different polarization SAR data to undertake specific applications.In this paper, the chosen test site is located in Fuzhou city in the southeast of China.Data obtained from C-band ERS1/2SAR, ENVISAT ASAR and L-band ALOS PALSARare processed into intensity image and interference phase information. Typical surfacefeatures of Fuzhou are analyzed and compared on backscatter and interference coherenceinformation between C and L-band SAR, and applied to land use/cover classification.The main research contents include:(1) SAR intensity imge processing and InSAR interference coherence processing. SARintensity imge processing steps include: data extraction, backscatter intensity imagegeneration, image matching, speckle filtering, radiometric calibration, geometriccorrection, geocoding, etc. InSAR processing steps include: SLC data acquisition, pluralimage co-registration, interferogram generation, the baseline estimation curved earthphase trend removal, interference coherence estimation, geocoding, etc. The main featuresin this paper are those:①Distinguished the difference among different mulut-looknumbers when multi-look processing;②Utilized auto-matching technology combinedwith DEM terrain data for geometric correction to improve the correction accuracy;③Image matched at fine-scale based on sub-pixel to improve the quality of resultinginterference coherence imge;④Analyzed multiple-scale coherent effects to obtain thebest interferogram. (2) Aanlysis of backscatter characteristics and coherence between C-band and L-bandSAR data. With muti-temporal C-band and L-band SAR data of Fuzhou area, backscatterand coherence properties of waters, residential areas, forest, paddy fields and other sitesof study area are analyzed and compared. It is founded that,①coherence coefficients ofwater are stable both in C-band and L-band, and backscatter values change with the watersurface roughness in C-band while stable in L-band;②backscatter and coherencecoefficient of residents both are the highest among all land cover types;③the value ofboth backscatter and coherence coefficient of forest in C-band are lower than that inL-band;④in the rice-growing season, the change of backscatter and coherencecoefficient of paddy fields is the greatest.(3) Land use/cover classification and comparison between C and L-band. Based onsimilarities and differences of backscatter and interference properties between C-band andL-band, intensity and coherence images ERS1/2SAR, ASAR and PALSAR arerespectively used for land use/cover classification. It is found that L-band data has betterresults in the extractions on residential areas, woodlands, waters and other types. In therice-growing season, C-band and L-band can both be used for rice extraction by usingtemporal differences of backscatter intensity and coherence coefficient acquired duringrice growth stages.