Saeid Taleghanidoozdoozan

Although Mamba models significantly improve hyperspectral image (HSI) classification, one critical challenge is the difficulty in building the sequence of Mamba tokens efficiently. This letter presents a Sparse Deformable Mamba (SDMamba) approach for enhanced HSI classification, with the following contributions. First, to enhance Mamba sequence, an efficient Sparse Deformable Sequencing (SDS) approach is designed to adaptively learn the ”optimal" sequence, leading to sparse and deformable Mamba sequence with increased detail preservation and decreased computations. Second, to boost spatial-spectral feature learning, based on SDS, a Sparse Deformable Spatial Mamba Module (SDSpaM) and a Sparse Deformable Spectral Mamba Module (SDSpeM) are designed for tailored modeling of the spatial information spectral information. Last, to improve the fusion of SDSpaM and SDSpeM, an attention based feature fusion approach is designed to integrate the outputs of the SDSpaM and SDSpeM. The proposed method is tested on three benchmark datasets with many state-of-the-art approaches, inlcuding CNNs, GAN Transformer, and Mamba based methods, demonstrating that the proposed approach can achieve higher accuracy with less computation, and better detail small-class preservation capability.

Processing SAR images for sea ice mapping using deep learning is typically constrained to analyzing neighboring pixels rather than studying visual patterns at a large scale. This quality ignores the wide within-class variance and spatial non-stationary class statistics observed by covering complete scenes. To leverage global and local information and improve automated sea ice mapping, we introduce the Irregular Tokens on Transformers (ITT) method by uniquely combining convolutional neural networks (CNN), transformers, and unsupervised segmentation. The method is trained to capture long-range spatial dependencies to overcome non-stationary class statistics. In addition, ITT incorporates a novel loss function, including a pixel-based term that guides the semantic segmentation and a region-based term that promotes region-consistent feature maps, increasing the classifier confidence and improving the predictions along object boundaries. We evaluate the method by classifying ice and water on dual-polarized RADARSAT-2 scenes. The evaluation metrics depict the best performance, generally, when the training leverages both pixel and region level loss terms, after pre-training the CNN. In such a case, prediction maps are less sensitive to non-stationary patterns, more detailed on ice/water boundaries, and less uncertain separating the classes.

Land cover classification from compact polarimetry (CP) imagery captured by the launched RADARSAT Constellation Mission (RCM) is important but challenging due to class signature ambiguity issues and speckle noise. This paper presents a new land cover classification method to improve the learning of discriminative features based on a novel pyramid fine- and coarse-grained self-attention transformer (PFC transformer). The fine-grained dependency inside a non-overlapping window and coarse-grained dependencies between non-overlapping windows are explicitly modeled and concatenated using a learnable linear function. This process is repeated in a hierarchical manner. Finally, the output of each stage of the proposed method is spatially reduced and concatenated to take advantage of both low- and high-level features. Two high-resolution (3 m) RCM CP SAR scenes are used to evaluate the performance of the proposed method and compare it to other state-of-the-art deep learning methods. The results show that the proposed approach achieves an overall accuracy of 93.63%, which was 4.83% higher than the best comparable method, demonstrating the effectiveness of the proposed approach for land cover classification from RCM CP SAR images.

Operational sea ice maps are usually generated manually using dual-polarization (DP) synthetic aperture radar (SAR) satellite imagery, but there is strong interest in automating this process. Recently launched satellites offer compact polarimetry (CP) imagery that provides more comprehensive polarimetric information compared to DP, which compels the use of CP for automated classification of SAR sea ice imagery. Existing sea ice scene classification algorithms using CP imagery rely on handcrafted features, while neural networks offer the potential of features that are more discriminating. We have developed a new and effective sea ice classification algorithm that leverages the nature of CP data. First, a residual-based convolutional neural network (ResCNN) is implemented to classify each pixel. In parallel, an unsupervised segmentation is performed to generate regions based on CP statistical properties. Regions are assigned a single class label by majority voting using the ResCNN output. For testing, quad-polarimetric (QP) SAR sea ice scenes from the RADARSAT Constellation Mission (RCM) are used, and QP, DP, CP, and reconstructed QP modes are compared for classification accuracy, while also comparing them to other classification approaches. Using CP achieves an overall accuracy of 96.86%, which is comparable to QP (97.16%), and higher than reconstructed QP and DP data by about 2% and 10%, respectively. The implemented algorithm using CP imagery provides an improved option for automated sea ice mapping.

Although the compact polarimetric (CP) synthetic aperture radar (SAR) mode of the RADARSAT Constellation Mission (RCM) offers new opportunities for oil spill candidate detection, there has not been an efficient machine learning model explicitly designed to utilize this new CP SAR data for improved detection. This paper presents a conditional random field model based on the Wishart mixture model (CRF-WMM) to detect oil spill candidates in CP SAR imagery. First, a “Wishart mixture model” (WMM) is designed as the unary potential in the CRF-WMM to address the class-dependent information of oil spill candidates and oil free water. Second, we introduce a new similarity measure based on CP statistics designed as a pairwise potential in the CRF-WMM model so that pixels with strong spatial connections have the same class label. Finally, we investigate three different optimization approaches to solve the resulting maximum a posterior (MAP) problem, namely iterated conditional modes (ICM), simulated annealing (SA), and graph cuts (GC). The results show that our proposed CRF-WMM model can delineate oil spill candidates better than the traditional CRF approaches, and that the GC algorithm provides the best optimization.