Science AI Journal
Engineering · Research topic

Open research questions in Remote-Sensing Image Classification

70 unresolved questions extracted from the limitations and future-work sections of 325 Remote-Sensing Image Classification papers in our library. Each links back to the study that raised it.

What the literature leaves open

  • Future studies can explore the application of the proposed model to other types of natural hazards. The study suggests that future research can focus on improving the model's performance using larger datasets. Future studies can investigate the use of other multimodal fusion strategies, such as early fusion or hybrid fusion.

    Post-hurricane building damage assessment using street-view imagery and structured data: a multimodal deep learning approach · 2026 · DOI

  • Existing post-hurricane building damage assessment methods have limitations, such as relying on satellite imagery data. There is a need for a multimodal approach that combines visual and contextual information. The study identifies a gap in the current literature, where most methods focus on single-modal data.

    Post-hurricane building damage assessment using street-view imagery and structured data: a multimodal deep learning approach · 2026 · DOI

  • Current LULC models are typically developed for a specific modality and a fixed class taxonomy, limiting their generalizability - There is a need for a universal model that can be applied to diverse LULC domains

    LandSegmenter: Towards a flexible foundation model for Land Use and Land Cover mapping · 2026 · DOI

  • Chaotic and sprawling destruction caused by major infrastructure failures. Limitations of optical sensors and SAR. Need for rapid and reliable change detection methods.

    A Multi-Sensor Semi-Supervised and Unsupervised Framework for Post-Disaster Flood and Building Damage Assessment: The Case of the Derna Dam Collapse · 2026 · DOI

  • Cloud cover and varying weather conditions. Capturing temporal change dynamics. Limited spatial resolution and computational intensity of existing methods.

    SAU-MTF: Siamese attention U-Net with multimodal temporal fusion for accurate deforestation detection · 2026 · DOI

  • Future research can focus on exploring the application of the proposed DCAMF framework to other computer vision tasks. Future research can investigate the use of other backbone encoders and architectures in the DCAMF framework. Future research can develop more efficient and effective methods for confidence-aware mask fusion.

    A Dynamic Confidence-Aware Mask Fusion Framework for Aerial Image Segmentation Using U-Net and U-Net++ · 2026 · DOI

  • Existing ensemble-based segmentation methods typically rely on homogeneous architectures or static weighting schemes. There is a gap in existing fusion strategies for combining heterogeneous models with diverse backbones. The proposed DCAMF framework addresses this gap by developing a mask-level fusion framework that leverages the strengths of U-Net and U-Net++ with diverse backbones.

    A Dynamic Confidence-Aware Mask Fusion Framework for Aerial Image Segmentation Using U-Net and U-Net++ · 2026 · DOI

  • The ability of SAR-to-optical image translation to reconstruct multispectral optical information is limited due to the single-polarization mode of SAR imagery. There is a need to improve the accuracy of S2O image translation using spatiotemporal features.

    Improving Generative Deep Learning-Based SAR-to-Optical Image Translation through Spatiotemporal Feature Incorporation: A Case Study in Cropland · 2026 · DOI

  • High computational complexity. Redundant spectral information. Limited labeled samples and class disparities.

    Central spectral-spatial context attention for hyperspectral image classification · 2026 · DOI

  • Scale-related challenges. Limited temporal coverage of observations. Imagery resolution limits the detection of fine morphological features.

    DeltaLatent: a latent-space framework for delta morphology classification · 2026 · DOI

  • The need for large annotated datasets. The complexity of urban scenes. The variability of building densities.

    GeoRGMAE: Geospatially Guided Masked Autoencoders for Building Segmentation · 2026 · DOI

  • Current methods have limitations in terms of accuracy and reliability. There is a need for a modular and multi-modal approach to change detection.

    PICANTEO: A Modular Change Detection Framework for Remote Sensing Applications · 2026 · DOI

  • While existing work has made efforts on these challenges individually, jointly improving fine-tuning performance under spectral mismatch while reducing adaptation cost remains underexplored.

    SPECTRA: Band-Routed Embedding and Stage-Wise LoRA for Cross-Sensor Fine-Tuning of Geospatial Foundation Models · 2026

  • Deep learning provides a promising solution for modeling complex spatiotemporal dependencies; however, its application is often limited by the difficulty of obtaining paired clear-sky and degraded NDVI data for identical spatiotemporal locations.

    Global-Scale Self-Supervised Spatiotemporal Learning for NDVI Time-Series Reconstruction · 2026

  • Recent advances in vision foundation models (VFMs) have demonstrated strong potential for diverse computer vision tasks; however, their practical applicability to high-resolution satellite imagery under operational constraints remains insufficiently explored.

    Evaluating the applicability of a vision foundation model with parameter-efficient adaptation for multitask analysis of KOMPSAT-3/3A satellite imagery · 2026 · DOI

  • Vision-language foundation models (FMs) provide a data-efficient paradigm through prompt-based semantic and spatial guidance, but the relative contribution of different prompt types remains unclear.

    Evaluating Semantic and Spatial Guidance for Foundation Model Segmentation of Small-Scale PV in Remote Sensing Imagery · 2026

  • Current deep learning-based methods face challenges in effectively extracting discriminative joint spatial-spectral features. Existing methods have limitations in modeling global relationships and capturing complex spatial contexts.

    Hyperspectral image classification network based on multiscale spatial-spectral fusion and semantic enhancement encoder · 2026 · DOI

  • This paper proposed a novel HSI classification network based on MSSF and SEE. The experimental results conducted on the Pavia University and Salinas datasets demonstrate that our approach effectively overcomes the limitations of existing methods in extracting discriminative joint spatial-spectral features. The proposed MSSF module successfully captures complementary information across multiple scales through spatial and spectral branches, enabling synergistic feature interaction that significantly enhances the characterization of complex land cover structures with highly variable spectral responses. Furthermore, the incorporation of the multi-head attention mechanism within the SEE module explicitly models global contextual relationships, thereby refining feature representation in semantically critical regions. The outstanding performance of MSNet is quantitatively validated, achieving remarkable OA of 95.68% on the Pavia University dataset and 96.84% on the Salinas dataset, which notably surpasses those obtained by existing mainstream methods. These results conclusively demonstrate the effectiveness and superiority of the proposed architectural design in advancing HSI classification performance. The methodology presented in this work provides a viable and effective solution for remote sensing image analysis tasks, offering significant potential for practical applications in environmental monitoring, agricultural management, and urban planning. Future research will focus on further extending the proposed framework to accommodate more diverse hyperspectral datasets acquired under different conditions, exploring strategies for enhancing computational efficiency to facilitate real-time processing of HSI.

    Hyperspectral image classification network based on multiscale spatial-spectral fusion and semantic enhancement encoder · 2026 · DOI

  • Prevailing superpixel-based methods face an inherent contradiction between clustering and classification. Conventional methods implement pixel-wise classification, leading to local inconsistencies and inaccurate edges.

    Hyperspectral image classification via efficient global spectral supertoken clustering · 2026 · DOI

  • Existing methods have limitations in capturing both local and global features. There is a need for a novel approach to integrate local and global features.

    Hierarchical fusion of local and global visual features with mixture-of-experts for remote sensing image scene classification · 2026 · DOI

  • Training data scarcity is a significant challenge, as DL models require thousands of annotated samples for effective training. Creating dense annotations requires expert photointerpretation combined with field validation, which is time-consuming and expensive. Historical field data for Rudrasagar are limited to specific studies, and systematic spatial sampling across the lake's area is lacking.

    Deep Learning Approaches for Environmental Monitoring of Wetlands Using Remote Sensing Data: Special Reference to Rudrasagar, a Ramsar Wetland of India · 2026 · DOI

  • Significant gaps exist in applying deep learning approaches to small, monsoon-influenced wetlands like Rudrasagar. The lack of systematic spatial sampling and limited historical field data hinder the development of effective monitoring models. The unique challenges of tropical Ramsar wetlands require innovative monitoring approaches.

    Deep Learning Approaches for Environmental Monitoring of Wetlands Using Remote Sensing Data: Special Reference to Rudrasagar, a Ramsar Wetland of India · 2026 · DOI

  • The lack of effective methods for monitoring U. prolifera outbreaks. The need for high-precision semantic segmentation of U. prolifera.

    A multi-attention and edge deep supervision neural network for capturing floating Ulva prolifera using GF-1 satellite imagery · 2026 · DOI

  • High computational cost and training stability issues. Limited scalability. Difficulty in handling high dimensionality and spectral redundancy in HSI data.

    Diffusion models for hyperspectral image analysis: A comprehensive review · 2026 · DOI

  • Investigating ways to improve the computational efficiency and training stability of diffusion models. Exploring applications of diffusion models in other domains. Developing new diffusion-based methods for HSI analysis.

    Diffusion models for hyperspectral image analysis: A comprehensive review · 2026 · DOI

Most-cited papers in Remote-Sensing Image Classification

Most recent work

Find a gap in your own Remote-Sensing Image Classification sub-topic

This page shows what the Remote-Sensing Image Classification literature already flags as unresolved. To narrow it to your specific question, run the guided finder — it searches the gap library on demand and checks candidates against 250M+ OpenAlex works.

Open the Research Gap Finder →

Related topics in Engineering

Traffic and Road Safety
237 open questions
Concrete and Cement Materials Research
196 open questions
Digital Transformation in Industry
190 open questions
Design Education and Practice
167 open questions
Machine Fault Diagnosis Techniques
165 open questions
Smart Cities and Technologies
152 open questions
Spatial Cognition and Navigation
149 open questions
BIM and Construction Integration
145 open questions
70 open questions have been extracted from the limitations and future-work passages of 325 Remote-Sensing Image Classification papers in our library. Each one below links back to the study that raised it, so you can read the original claim in context.

Tools for your next paper

Gaps & Literature SearchFind open questions in any field — 120,000+ open gaps mined from our own library.Pre-CheckIs your paper ready? A Tier 1–5 acceptance probability in seconds.Journal FinderA ranked shortlist of target journals from a 17,500-venue index.Duplicate Publication CheckerCheck prior-publication & salami-slicing risk across 8 sources.Citation GeneratorFreeA DOI or title → APA, MLA, Vancouver, Chicago, BibTeX, RIS. No signup.Graphical Abstract MakerFreeTurn your findings into a colorblind-safe graphical abstract.

Compare the category — Honest roundups of the AI research tools, ours listed alongside the alternatives.

Command palette

Jump anywhere, run any action.