Kai Sun

Seismic interpretation for three-dimensional (3D) subsurface structural modeling requires detected fault samples to be assigned to individual faults for inspection and surface construction. Intersections and fragmented observations make these assignments difficult, and automatic geometric partitions can contain incorrect merges and splits. We propose Consensus Relation Supervision (CRS) for 3D seismic fault instance segmentation using two uncorrected geometric partitions. Spatial overlap matching extracts strict consensus cores, whose same-fault and different-fault sample pairs provide supervision independent of numerical instance identifiers. A 3D-UNet3+ network learns instance embeddings and auxiliary separation cues from binary fault volumes. Point-pair training combines local and long-range relations, with local sampling prioritizing crossing negatives and positives at weak connections. The composite objective combines a balanced soft-target relation loss with hard-positive consistency, crossing-negative separation, and auxiliary boundary supervision. At inference, learned relations guide graph grouping and label propagation, followed by geometric association of compatible fragments. Manually revised instance labels are reserved for validation, parameter selection, and evaluation. Synthetic comparisons support improved instance grouping with the selected relation targets, and qualitative applications to F3D and A2 show fault-instance organization on field data. By learning association and separation from noisy automatic interpretations, CRS produces sample groups for inspecting individual faults and constructing their surfaces, providing an instance representation for subsequent 3D subsurface structural modeling.

Automatic fault interpretation is essential for seismic geological modeling; however, fault intersections, noise interference, and topological complexity frequently lead to over-segmentation or under-segmentation. To address these challenges, we propose multi-scale contour constraints Delaunay-based region growing (MC-DBRG), an instance segmentation method for three-dimensional (3D) fault point clouds. This approach effectively extracts individual fault instances from fault probability volumes. The pipeline begins with voxel downsampling and multi-scale feature extraction applied to the fault point cloud. A Gaussian Mixture Model then performs binary classification to delineate explicit and closed fault contour walls. These walls function as physical barriers during the subsequent Delaunay triangulation-based region growing, which initializes from non-contour smooth points. Implementing these constraints guides the growth process and prevents topological adhesion among intersecting faults. To correct over-segmentation caused by data discontinuities in the preliminary results, we introduce a global manifold secondary geometric merging strategy. A macroscopic planar trend constraint is subsequently applied to accurately reassign and backfill contour points and noise. Experimental results indicate that MC-DBRG resolves boundary ambiguity and spatial adhesion within complex intersecting faults. The generated fault models preserve high topological integrity and precise instance labels, establishing a reliable foundation for automated seismic geological modeling.

While deep learning methods have achieved significant success in pan-sharpening, existing cross-modal upsampling methods typically adopt a premature fusion strategy, causing low frequency and high-frequency components to be coupled within the mixed features. To address this frequency coupling problem, we propose the Cross-domain Wavelet Frequency Upsampling (CWFU) method. Specifically, CWFU adopts a decompose-at-source strategy: leveraging wavelet transform to separate PAN into low-frequency global structures and high-frequency local details before cross-modal fusion, establishing explicit frequency boundaries. Low-frequency components are processed in the Fourier domain by the Low-Frequency Feature Fusion module, achieving robust fusion of global structures through adaptive amplitude gating and phase convolution while avoiding high-frequency interference. Subsequently, in the spatial domain, the High-Frequency Detail Injection module injects fine-grained details into high-frequency components, guided by the fused low-frequency features to eliminate low-frequency interference. These enhanced components are then reconstructed through inverse wavelet transform, achieving synergistic enhancement of global and local information. Furthermore, as a lightweight plug-and-play module, CWFU can be seamlessly integrated into various backbone networks. Extensive experiments on four benchmark datasets demonstrate the effectiveness of our CWFU, which consistently outperforms state-of-the-art upsampling methods.

While deep learning methods have demonstrated significant success in pan-sharpening, most existing approaches suffer from insufficient high-frequency guidance from panchromatic (PAN) images during the upsampling of low-resolution multi-spectral (LRMS) images, often failing to reconstruct fine-grained local details and maintain global contextual consistency. To alleviate these limitations, we propose a novel Detail-aware local and Context-aware global Dual-branch Upsampling (DCDU) method for pan-sharpening. The core of DCDU is a dual-branch architecture specifically designed to decouple and leverage both local structural details and global semantic information from both modalities: the detail-aware local branch calculates feature-space correlations between LRMS images and PAN images to derive adaptive kernels, guiding the reconstruction of local neighborhoods and facilitating high-fidelity injection of fine-grained spatial structures; the context-aware global branch mainly employs a cross-modality channel-specific spatial attention mechanism to holistically model long-range dependencies, dynamically re-calibrating spectral-spatial representations based on cross-modal context and capturing macro-level semantic coherence across the entire scene. The distinct information streams from the two branches are then integrated via an adaptive and learnable mapping function. We also provide a theoretical analysis under the minimum mean square error criterion to motivate our dual-stream strategy, the efficacy of which is robustly validated by our experimental results. As a generic and lightweight upsampling module, the collaborative design can be seamlessly integrated into various backbone networks, enabling superior performance across multiple datasets.

Pansharpening, a critical task in remote sensing, aims to fuse panchromatic (PAN) images with low-resolution multispectral (LRMS) images to generate high-resolution multispectral (HRMS) outputs. Existing deep learning methods primarily focus on pixel-level spatial and spectral features, often overlooking the deeper, text-level semantic information that can enhance fusion. To address this limitation, we introduce a novel Text-modulated Multiscale Guidance Transformer (TMGformer). Specifically, we first establish a self-contained workflow to generate text features, eliminating reliance on external Application Programming Interfaces (APIs). Subsequently, we develop a Text-modulated Visual Feature Refinement (TVFR) module that transforms these text features into spectrally compatible and semantically consistent guidance. This guidance is then injected into our Asymmetric Multiscale Fusion (AMF) module to guide the fusion process at multiple scales, effectively addressing the vast variance in the scale of terrestrial objects described in their corresponding texts. Furthermore, we introduce a hybrid text-visual loss function to provide comprehensive supervision. Extensive experiments demonstrate that TMGformer achieves state-of-the-art performance. Notably, on the GaoFen-2 dataset, our method reduces the ERGAS by 5.11% and the D_s by 16.44% compared to the second-best competitor. To facilitate future research, we release a new vision-language dataset containing textual descriptions and corresponding text features. The code and dataset are available at: https://github.com/PeterZhaoXJTU/TMGformer.

Lane detection is a critical and challenging task in autonomous driving, particularly in real-world scenarios where traffic lanes can be slender, lengthy, and often obscured by other vehicles, complicating detection efforts. Existing anchor-based methods typically rely on prior lane anchors to extract features and subsequently refine the location and shape of lanes. While these methods achieve high performance, manually setting prior anchors is cumbersome, and ensuring sufficient coverage across diverse datasets often requires a large amount of dense anchors. Furthermore, the use of Non-Maximum Suppression (NMS) to eliminate redundant predictions complicates real-world deployment and may under-perform in complex scenarios. In this paper, we propose Polar R-CNN, an end-to-end anchor-based method for lane detection. By incorporating both local and global polar coordinate systems, Polar R-CNN facilitates flexible anchor proposals and significantly reduces the number of anchors required without compromising performance. Additionally, we introduce a triplet head with heuristic structure that supports NMS-free paradigm, enhancing deployment efficiency and performance in scenarios with dense lanes. Our method achieves competitive results on five popular lane detection benchmarks—Tusimple, CULane, LLAMAS, CurveLanes, and DL-Rail—while maintaining a lightweight design and straightforward structure. Our source code is available at https://github.com/ShqWW/PolarRCNN

In the field of remote sensing (RS) imaging, the occurrence of adverse meteorological conditions or sensor malfunctions can lead to missing data, posing a substantial impediment. Low-rank tensor decomposition has emerged as a promising strategy for resolving this issue, as it enables the integration of diverse data priors within a unified framework. Although various decomposition techniques, such as Tucker decomposition and tensor ring decomposition, have been developed based on multi-linear products, they may not adequately capture the complex structure of RS imagery. Therefore, there is a need for tensor decompositions that incorporate nonlinear operations. To alleviate this challenge, a multi-nonlinear product is defined, which enables the construction of a nonlinear Tucker decomposition model. To enhance the model’s capability, a stacked Tucker decomposition model is formulated, by representing a tensor as the product of a core tensor and a collection of factor matrices along each mode, utilizing the multi-nonlinear product, which potentially regulates the distribution of singular values, thereby achieving a more accurate characterization of textures. The proposed model, integrated with total variation regularization, is subsequently applied to the task of RS imagery inpainting. Extensive experimental results demonstrate the superiority of the proposed model over state-of-the-art methods across various tasks. This validates its effectiveness and adaptability in mitigating the challenges associated with RS imagery inpainting. The code is available at https://github.com/shuangxu96/STDTV.