Multifrequency Spaceborne Synthetic Aperture Radar Data for Backscatter-Based Characterization of Land Use and Land Cover

Abstract

Polarimetric synthetic aperture radar remote sensing extracts the information about the target using decomposition models to separate the polarimetric information into single-bounce (contributed by smooth surfaces), double-bounce (contributed by urban structure), and volume (mainly due to vegetation cover) scattering components. The penetration capacity of the electromagnetic wave into the surface increases with the decrease in its frequency. This study explores and compares the polarimetric decomposition models for scattering-based characterization of land use and cover using multifrequency spaceborne synthetic aperture radar sensor datasets that were acquired over San Francisco, CA, USA. The present work compares the scattering parameters of coherent (Pauli), roll-invariant (Barnes), eigenvalue–eigenvector (Cloude), and compact-polarimetric (Raney) decomposition modeling approaches for scattering-based characterization of urban structures, waterbody, and vegetation cover. The land use/cover classification was performed based on the scattering response of the scatterers using a support vector machine classifier. The outputs of the classification approach on multisensor, multifrequency, and multi-polarization polarimetric synthetic aperture radar data have shown reasonable accuracy in classifying the land use and land cover. The decomposition models fail to characterize the oriented urban structures that cause misclassification of urban structures as vegetation. The higher-order roll-invariant decomposition modeling approaches could improve the interpretation of different targets and accuracy in land use and land cover classification.

Polarimetric Decomposition and Machine Learning‐Based Classification of L‐ and S‐Band Airborne SAR (LS‐ASAR) Data

“…This misidentification in the output of the decomposition model was due to the volumetric scattering obtained from oriented buildings that were very much similar to the volumetric scattering obtained from the vegetation. Therefore, similar to previous studies (Garg et al., 2022; Jafari et al., 2015; Verma et al., 2022), the output classification image showed a misclassification of oriented urban buildings into vegetation class (Figure 8). The classification of oriented buildings could be improved by utilizing interferometric polarimetry components—coherence amplitude and interferogram, which reduces the confusion between urban and vegetation pixels (Qi et al., 2012).…”

Accuracy Assessment of different classifiers for Sustainable Development in Landuse and Landcover mapping using Sentinel SAR and Landsat-8 data

“…Although the input training data provided in this study may not be sufficient for some classifiers, KD tree KNN and Minimum distance to mean classifiers require more training datasets to acquire results efficiently [33]. In a random forest classifier, the number of classes that were specified for the given input training data has improved the efficiency and correlation of the classifier, producing good results [27]. When compared to other classifiers, the random forest classifier produces accurate results [28].…”

Dielectric characterization and polarimetric analysis of lunar north polar crater Hermite-A using Chandrayaan-1 Mini-SAR, Lunar Reconnaissance Orbiter (LRO) Mini-RF, and Chandrayaan-2 DFSAR data

“…SAR data can be used to study the geophysical parameters of the lunar surface and it can be done with help of polarimetric data-based scattering properties retrieval of the lunar surface (Vashishtha & Kumar, 2021;. Polarimetric SAR (PolSAR) can differentiate scattering elements in a single resolution cell (Babu et al, 2022;Garg et al, 2022;Maiti et al, 2021;Verma et al, 2022). A target area is an area that contributes to three types of scattering patterns namely Surface scattering, double-bounce scattering, and volume scattering and the mixture of these scattering patterns gives precise details about the physical properties of the target area (Kumar et al, 2019(Kumar et al, , 2020(Kumar et al, , 2022b.…”