Search

The polarimetric Synthetic Aperture Radar (SAR) data sets have been widely exploited for land use land cover (LULC) classification due to their sensitivity to the structural and dielectric properties of the imaging target. In this study, the potential of fully polarimetric L‐ and S‐band Airborne SAR (LS‐ASAR) data sets were explored for the machine‐learning‐based classification of Urban, Vegetation, Waterbody, and Open Ground. This work was done by utilizing dual‐frequency L‐ and S‐band airborne data of Santa Barbara, California, USA, acquired under the Airborne SAR (LS ASAR) campaign, a precursor airborne mission to the space‐borne NASA‐ISRO (NISAR) mission. The LS‐ASAR polarimetric information was utilized for LULC classification using the SVM classifier. The roll‐invariant Barnes, and eigenvalue/eigenvector‐based Cloude and H/A/Alpha decomposition were implemented to retrieve the scattering parameters. The backscatter response of classes was studied, and separability analysis was done to reduce the misclassification error between six class pairs‐ Vegetation—Urban, Vegetation—Waterbody, Vegetation—Open Ground, Urban—Waterbody, Urban—Open Ground, and Water—Open Ground. The decomposition models failed to achieve the desirable separability index for all six class pairs; consequently, the classification of Barnes, Cloude, H/A/Alpha decomposition showed misclassification between vegetation‐urban class, and waterbody‐open ground class for both L‐ and S‐band data sets. The effort was made toward improving the classification accuracy by integrating the roll‐invariant and eigenvalue‐eigenvector scattering parameters of the multifrequency L‐ and S‐band data set. This method presented the desirable separability index for all class‐pair; eventually highest classification accuracy was achieved i.e. 93.35% (kc ${k}_{c}$ = 0.91) by significantly reducing the misclassification error between class pairs. Key Points: Implementation of Barnes, Cloude & H/A/Alpha polarimetric decomposition models on dual‐frequency polarimetry SAR L‐ and S‐band data setsSupport Vector Machine classification of scattering parameters and generation of confusion matrices for Accuracy AssessmentImproving the classification accuracy by integrating the polarimetric scattering components of L‐ and S‐band data sets [ABSTRACT FROM AUTHOR]

The Polarimetric Synthetic Aperture Radar (PolSAR) systems use electromagnetic radiations of different polarizations in the microwave frequency to collect the scattering information from targets on the Earth. Nevertheless, as with any other electronic device, the PolSAR systems are also not ideal and subjected to distortions. The most important of these distortions are the polarimetric distortions caused due to the channel imbalance, phase bias, and crosstalk between the different polarization channels. For the spaceborne PolSAR systems, the Earth's ionosphere also contributes to an additional polarimetric distortion known as the Faraday rotation. An effort was made in this study to perform the polarimetric calibration of the Quad-pol and Compact-pol PolSAR datasets acquired using different airborne and spaceborne PolSAR systems to estimate and minimize these polarimetric distortions. The investigation was also done to analyze the impact of these polarimetric distortions on the scattering mechanisms from ground targets and on its dependency on the radar wavelength. The study was done using the UAVSAR L-band Quad-pol dataset, RADARSAT-2 Quad-pol dataset, ALOS-2 PALSAR-2, ISRO's L&S- Band Airborne SAR (LS-ASAR) Quad-pol and Compact-pol datasets, and the RISAT-1 Compact-pol dataset. Calibration of the airborne PolSAR data was carried to understand the level of polarimetric distortions in the LS-ASAR product that is a precursor mission to the spaceborne Dual-Frequency L&S Band NASA-ISRO Synthetic Aperture Radar (NISAR) mission. It is understood that the crosstalk is the dominant polarimetric distortion, which severely affects the PolSAR datasets compared to the other polarimetric distortions, and it is more for the higher wavelength PolSAR systems. The Quegan, Improved Quegan, and Ainsworth algorithms for crosstalk estimation and minimization was performed for the different Quad-pol datasets and it was found that the Improved Quegan algorithm is suitable for removing crosstalk from datasets having high crosstalk and the Ainsworth algorithm is suitable for removing crosstalk from datasets having low crosstalk. The Freeman method of the polarimetric calibration was implemented for the compact-pol datasets and it was able to considerably minimize the polarimetric distortions. The coherency matrix, scattering matrix, model-based decomposition, polarimetric signatures, and roll invariant parameter-based analysis revealed that all the datasets after polarimetric calibration were showing the correct scattering responses expected from the ground targets. [ABSTRACT FROM AUTHOR]

This study compares the utility of multifrequency SAR and Optical multispectral data for land-cover classification of Mumbai city and its nearby regions with a special focus on water body mapping. The L-band ALOS-2 PALSAR-2, X-band TerraSAR-X, C-band RISAT-1, and Sentinel-2 datasets have been used in this work. This work is done as a retrospective study for the dual-frequency L and S-band NASA-ISRO Synthetic Aperture Radar (NISAR) mission. The ALOS-2 PALSAR-2 data has been pre-processed before the implementation of machine learning algorithms for image segmentation. Multi-looking is performed on ALOS-2 PALSAR-2 data to generate square pixels of size 5.78 m and then target decomposition is applied to generate a false-color composite RGB image. While in the case of TerraSAR-X and RISAT-1 datasets, no multi-looking was performed and direct target decomposition was applied to generate false-color composite RGB images. Similarly, for the optical dataset that has a resolution of 10 m, a true color composite, and a false color composite RGB image are generated. For the comparative study between ALOS-2 PALSAR-2 and Sentinel-2 dataset, the RGB images are divided into smaller chunks of size 500*500 pixels each to create a training and testing dataset. Ten image patches were taken from the large dataset, out of which eight patches were used to train the machine learning models Random Forest (RF), K Nearest Neighbor (KNN), and Support Vector Machine (SVM), and two patches were kept for testing and validation purpose. For training the machine learning models, feature vectors are generated using the Gabor filter, Scharr filter, Gaussian filter, and Median filter. For patch 1, the mIOU for true-color composite based Optical image varies from 0.2323 to 0.2866 with the RF classifier performing the best and the mIOU for false-color composite based Optical image varies from 0.4130 to 0.4941 with the RF classifier performing the best while for ALOS-2 PALSAR-2 data, the mIOU varies from 0.4033 to 0.4663 with the RF classifier outperforming the KNN and the SVM classifiers. For patch 2, the mIOU for true-color composite based Optical data varies from 0.3451 to 0.4517 with KNN performing the best and the mIOU for false-color composite based Optical image varies from 0.5156 to 0.5832 with the RF classifier performing the best while for ALOS-2 PALSAR-2 data, the mIOU varies from 0.4600 to 0.5178 with the RF classifier outperforming the KNN and the SVM classifiers. The gap between the performance of ALOS-2 PALSAR-2 data and Sentinel-2 optical data is observed when the IOU of water class is compared, with IOU w for the true-color composite based optical image at a maximum of 0.2525 and for false-color composite based optical image at a maximum of 0.7366 while for ALOS-2 PALSAR-2 data a maximum IOU w of 0.7948 is achieved. The better performance of SAR data as compared to true-color composite based optical image data is due to the misclassification of ground and water classes into urban and forest in the case of the true-color composite based optical dataset which can be attributed to the high similarity between water and forest classes in the case of true-color composite based optical data whereas both these classes are easily separable in case of SAR data. This issue is however resolved by using the false color composite based optical image dataset for the classification task which performs slightly better than ALOS-2 PALSAR-2 data in the overall classification task. However, the SAR data works best in water body detection as notable from the high IOU for water class in the case of SAR data. In addition to the comparative analysis between Sentinel-2 optical and ALOS-2 PALSAR-2 data, land-cover classification has been performed on X-band TerraSAR-X and C-band RISAT-1 data on a single patch and it has been found that the RF classifier performs the best, recording the mIOU 0.5815 for X-band TerraSAR-X data, mIOU of 0.4031 for the C-band RISAT-1 data, and mIOU of 0.6153 for the L-band ALOS-2 data. [ABSTRACT FROM AUTHOR]