Semantic segmentation is crucial for a wide range of downstream applications in remote sensing, aiming to classify pixels in remote sensing images (RSIs) at the semantic level. The dramatic variations in grayscale and the stacking of categories within RSIs lead to unstable inter-class variance and exacerbate the uncertainty around category boundaries. However, existing methods typically emphasize spatial information while overlooking frequency insights, making it difficult to achieve desirable results. To address these challenges, we propose a novel dual-domain fusion network that integrates both spatial and frequency features. For grayscale variations, a multi-level wavelet frequency decomposition module (MWFD) is introduced to extract and integrate multi-level frequency features to enhance the distinctiveness between spatially similar categories. To mitigate the uncertainty of boundaries, a type-2 fuzzy spatial constraint module (T2FSC) is proposed to achieve flexible higher-order fuzzy modeling to adaptively constrain the boundary features in the spatial by constructing upper and lower membership functions. Furthermore, a dual-domain feature fusion (DFF) module bridges the semantic gap between the frequency and spatial features, effectively realizes semantic alignment and feature fusion between the dual domains, which further improves the accuracy of segmentation results. We conduct comprehensive experiments and extensive ablation studies on three well-known datasets: Vaihingen, Potsdam, and GID. In these three datasets, our method achieved 74.56%, 73.60%, and 81.01% mIoU, respectively. Quantitative and qualitative results demonstrate that the proposed method significantly outperforms state-of-the-art methods, achieving an excellent balance between segmentation accuracy and computational overhead.

Hyperspectral unmixing (HU) aims to decompose mixed pixels into endmembers with corresponding abundances. However, while several existing convolutional autoencoder methods usually use fixed convolutional kernels, making it difficult to capture the global context. In addition, due to the huge solution space of unmixing, existing methods usually adopt a consistent sparsity constraint and lack adaptivity. To overcome the above limitations, we propose a multi-scale spectral-spatial unmixing network with Boltzmann-inspired adaptive temperature (MSSTNet). First, the spectral attention block (SEAB) and spatial attention block (SAAB) are designed to capture the dependence between spectral bands and enhance spatial feature extraction, respectively. These are integrated into multi-scale spectral-spatial attention blocks (MSSABs) with varying convolution kernels, which enable the network to focus on local and global image structures at the same time. Moreover, inspired by the Boltzmann distribution, we introduce a temperature matrix T in the softmax activation to regulate the output sparsity, similar to the effect of temperature on the particle energy distribution. The Euclidean distance and cosine distance between adjacent pixels are used to construct the similarity matrix to capture the spectral difference caused by the amplitude change, and then the T matrix is constructed. The softmax layer is divided by the resulting T matrix, so as to impose sparsity constraints of varying strengths on different areas. Evaluations on simulated and real datasets demonstrate the proposed approach's superiority over state-of-the-art methods. The code is publicly available at https://github.com/RSstudy/MSSTNet.

In remote sensing images, accurate semantic segmentation faces significant challenges due to the variation of object scales, uncertain category boundaries, and complex scenes. In view of the above challenges, we propose a prior-guided fuzzy-aware multibranch network for remote sensing image segmentation. Specifically, a prior-feature extractor (PFE) is designed to take the local features extracted by convolution structure as prior knowledge of the network. Fuzzy-aware module (FAM) is presented to perceive and refine category boundaries with fuzzy learning, which transforms the uncertainty problem into a quantitative analysis problem by establishing fuzzy relationships between neighborhood pixels. Multibranch-feature extractor (MFE) is put forward to aggregate multiscale global context information by combining positional attention and transformer. The whole network learning process is supervised by multibranch loss. We validated our method on the GID, Potsdam and LoveDA datasets, achieved 75.58%, 75.62% and 53.10% mean intersection over union (mIoU), respectively, which demonstrated the superiority of our method. In addition, ablation studies further demonstrate the validity of each module in the proposed method.

The increasing availability of high-resolution remote sensing images (RSIs), driven by advancements in satellite and optical imaging technologies, presents both new opportunities and challenges for small object segmentation (SOS) tasks. A primary challenge in SOS tasks is the substantial sample imbalance between foreground and background, resulting in the under-representation or even loss of features for small foreground targets. Furthermore, grayscale variability intensifies boundary uncertainty, complicating accurate segmentation. To address these issues, we propose a fuzzy visual perception network (FVPNet), which incorporates a feature inheritance downsampling (FID) module, a fuzzy refinement (FR) subnet and a small target perception (STP) subnet, along with a hierarchical balance loss (HB-Loss) that prioritizes small foreground objects to effectively reduce foreground bias in regression. The FID module integrates diverse feature map extraction techniques to construct a more robust feature map through various subsampling methods, mitigating feature loss in foreground targets. The FR subnet employs high-order fuzzy modeling with adaptive membership functions and a constrained-space boundary adaptation strategy to address boundary uncertainty in RSIs. Meanwhile, the STP subnet adopts a “coarse-to-fine” paradigm, utilizing a pyramidal structure alongside an auxiliary semantic decoder to enhance the network's capacity for accurate small target localization. We conduct extensive experiments and ablation studies on three benchmark datasets, iSAID, Vaihingen and Potsdam datasets. Across the three datasets, our method achieved mean Intersection over Union (mIoU) scores of 73.08%, 87.37%, and 81.54%, respectively. Both quantitative and qualitative evaluations clearly demonstrate that the proposed approach significantly outperforms existing state-of-the-art methods, achieving a superior balance between segmentation accuracy and computational efficiency.