MMMengru Ma

Mengru Ma

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Mengru Ma has a total of 41 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Modality-Prior Aware Progressive Diffusion Pansharpening

    by Jiaxuan Zhao, Lingling Li, Licheng Jiao, Mengru Ma, Xu Liu, Chao Wang, Fang Liu, Wenping Ma, Shuyuan Yang

    Pansharpening entails significant challenges in fully exploiting the synergistic potential between generative expressiveness and discriminative precision, requiring advanced generative modeling guided by discriminative priors. To this end, a unified end-to-end, modality-prior aware progressive diffusion model (PPDM-Pan) is presented that dynamically integrates discriminative priors with generative representations through a progressive fusion process, achieving synergistic feature interaction and enhanced spatial–spectral consistency for high-fidelity high-resolution multispectral reconstruction. Specifically, to explicitly extract modality-specific discriminative priors, a modal prior guidance (MPG) encoder is designed to models complementary spatial and spectral priors from heterogeneous panchromatic (PAN) and low-resolution multispectral (LRMS) inputs, establishing robust modality-aware representations. Subsequently, a diffusion process is leveraged to model the complex, high-dimensional PAN-LRMS distribution and synthesize expressive latent features with enhanced generative capacity. Crucially, to orchestrate synergistic integration of generative features and discriminative priors, a progressive modality-aware fusion (PMAF) strategy is introduced that hierarchically refines multi-level representations through content-aware dynamic learning, reconciling their inherent trade-offs to deliver high-fidelity reconstruction with superior spatial-spectral consistency. Experiments on QuickBird, GaoFen-2, and WorldView-3 datasets demonstrate its superior performance in balancing spatial realism, characterized by visually realistic textures and structural details, and spectral fidelity, highlighting its effectiveness as a robust and versatile solution.

  • HCMA-Net: Hierarchical Cross-Modality Aggregation Network for Multimodal Remote Sensing Image Classification

    by Wenping Ma, Hekai Zhang, Mengru Ma, Boyou Xue, Hao Zhu

    While multimodal remote sensing images provide complementary information in different imaging ways, effectively aggregating and jointly learning from these heterogeneous features is non-trivial, primarily due to the inherent modality gap that hinders semantic alignment and feature interoperability. To overcome these issues, we propose a Hierarchical Cross-Modality Aggregation Network (HCMA-Net). It introduces two novel components: the hierarchical feature aggregation and enhancement module (HFAE-Module) and the cross-modality interactive feature extraction module (CMIFE-Module). The HFAE-Module tackles the modality gap and enables cross-scale interaction through its Hierarchical Cross-Modality Feature Aggregation (HCMFA) mechanism, which incorporates Cross-Spectral and Spatial Aggregation Non-Local Attention layers (CSNLA and SANLA) to align features and aggregate contextual information across spectral and spatial dimensions. The CMIFE-Module addresses the optimization conflict by leveraging a dual-attention design; it uses self-attention to reinforce intra-modal coherence and cross-attention to dynamically extract and fuse complementary inter-modal features, thereby maximizing complementarity while avoiding the dilution of discriminative features and preventing negative transfer. Experiments on four real-world datasets (Hohhot, Nanjing, Xi’an, Houston2013) demonstrate that HCMA-Net consistently achieves outstanding classification results. The code is available at: https://github.com/sun740936222/HCMA-Net.

  • A Diff-Attention Aware State-Space Fusion Model for Remote Sensing Classification

    by Wenping Ma, Boyou Xue, Mengru Ma, Chuang Chen, Hekai Zhang, Hao Zhu

    Multispectral (MS) and panchromatic (PAN) images describe the same land surface, so these images not only have their own advantages, but also share a significant amount of redundant information. In order to separate similar information and each modality’s unique advantages, thereby reducing feature redundancy at the fusion stage, this paper introduces a diff-attention aware state space fusion model (DASF-Model) for multimodal remote sensing image classification. Based on the selective state space model, a cross-modal diff-attention module (CDAM) is designed to extract and separate the common features and their respective dominant features of MS and PAN images. Specifically, space preserving visual mamba (SPVM) retains image spatial features and captures local features by appropriately optimizing visual mamba’s input. Considering that features in the fusion stage will have large semantic differences after feature separation and traditional mean fusion method fails to effectively integrate these features with significant discrepancies, an attention-aware linear fusion module (ALFM) is proposed. It performs pixel-wise linear fusion by calculating influence coefficients. This mechanism can fuse features with large semantic differences while keeping the feature size unchanged. Empirical evaluations indicate that the presented method achieves better results than alternative approaches. The relevant code can be found at: https://github.com/AVKSKVL/DAS-F-Model.


  • Dense-Weak Ship Detection Based on Foreground-Guided Background Generation Network in SAR Images

    by Wenping Ma, Xiaoting Yang, Hao Zhu, Xiaoteng Wang, Biao Hou, Mengru Ma, Yue Wu

    Currently, ship detection based on synthetic aperture radar (SAR) images still faces significant challenges, particularly in detecting weak and densely distributed ships within complex backgrounds. In areas such as ports and land, the complex background features often resemble those of densely distributed ships, leading to reduced detection accuracy. In addition, the overlapping and mutual interference of features among dense ships can cause the network to miss detections or produce false positives. Therefore, this article proposes a foreground-guided background generation network (FGBG-Net), which includes a Gaussian foreground localization (GFL) model and a background feature removal (BFR) module. The GFL module identifies the approximate high-probability regions of ship foregrounds on the feature map, guiding the network to focus on these regions. The BFR module then progressively removes background interference features based on the positions provided by the GFL module, generating feature maps that are more suitable for detecting weak and dense ships. Our network has been validated on multiple SAR ship datasets, and the experimental results demonstrate noticeable performance improvements, with a mean average precision (mAP) increase of 3.4% on the SSDD and HRSID datasets. The relevant code is available at the following link: https://github.com/Xidian-AIGroup190726/FBGBNet/tree/master


  • Dual-Path Prototype Feature Decoupling Alignment Network for Panchromatic and Multispectral Classification

    by Wenping Ma, Yanshan Guo, Hao Zhu, Wenhao Zhao, Mengru Ma, Yue Wu, Licheng Jiao

    In recent years, with the rapid advancements and widespread application of satellite photography technology, it has become increasingly possible to obtain high-quality panchromatic (PAN) and multispectral (MS) data, which has provided new opportunities and challenges for multi-source information fusion and classification research. Remote sensing data has the characteristics of small inter-class differences and large intra-class differences, which easily leads to category confusion in network learning. In addition, how to fully tap the advantages of multi-source data, better align multi-source features, improve classification accuracy, and achieve collaborative classification are key issues that need to be solved urgently. In this paper, a dual-path prototype feature decoupling alignment network (DPFDA-Net) is designed to solve the above issues. The network consists of two components: a prototype feature embedding module (PFE) and a feature alignment module (FAM) based on prototype decoupling. In the feature extraction stage, the PFE module uses the prototype concept to learn the discriminative prototype features of each category of the dual-source data separately, making the boundaries between categories more obvious. The FAM operates at the dual-source prototype feature level and achieves feature alignment by decoupling single-source prototype features and performing feature transformation to supplement the missing information of another data source. Finally, we use the aligned features for classification. The results of the experiment demonstrate that our approach has made significant progress in improving classification precision. The code is available at https://github.com/Xidian-AIGroup190726/DPFDANet.


  • A 3D Self-Awareness Diffusion Network for Multimodal Classification

    by Mengru Ma, Wenping Ma, Licheng Jiao, Lingling Li, Xu Liu, Fang Liu, Shuyuan Yang, Yuwei Guo

    As imaging sensor technology in remote sensing has advanced quickly, multimodal fusion classification has become an important research direction in land cover and urban planning classification tasks. While generative models and image classification have greatly benefited from diffusion models, the present ones primarily concentrate on single-modality-driven diffusion processes. Therefore, this paper presents a 3D self-awareness diffusion network (3DSA-DiffNet) for multispectral (MS) and panchromatic (PAN) image fusion classification, which would make it easier to classify heterogeneous data from various sensors. First, in order to model the relationship between multi-channel spectra and multi-pixel spatial distributions as well as samples, respectively, a spatial-spectral joint denoising network (S $^{2}$ JD-Net) is proposed. It can incorporate the diffusion process into the neural network to enhance the quality of diffusion features. Secondly, to imitate the brain's spatial-spectral coexistence learning mechanism, this work offers a 3D self-awareness module (3DSA-Module) that can learn the weight of each pixel in 3D space, resulting in extraordinarily high feature representation capabilities. Finally, experimental verification demonstrates that the 3D self-awareness diffusion fusion network driven by brain inspiration outperforms more sophisticated approaches on the Xi'an, Huhhot, and Muufl datasets.


  • An Adaptive Dual-Supervised Cross-Deep Dependency Network for Pixel-Wise Classification

    by Wenping Ma, Chuang Chen, Mengru Ma, Hekai Zhang, Hao Zhu, Licheng Jiao

    With the advancement of remote sensing technology and satellite observation, the task of fusing multi-source data, such as multispectral (MS) and panchromatic (PAN) images, has become increasingly important. However, image fusion involving certain semantic differences can hinder the model’s ability to learn effective feature mappings. To reconstruct richer and more consistent features during fusion, we propose an adaptive dual-supervised cross-deep dependency network (ADCD-Net), which consists of two training stages. Stage I uses a semantic perceptual self-supervision strategy (SPS) to learn deep features across different modalities, thereby reducing semantic differences while mining its own non-singular features. Stage II uses the deep temporal Mamba module (DTM-Moudule) to interactively learn the output of each network layers, which are able to take part in the deep feature reinforcement and improve the classification performance of semantic information. Finally, to eliminate channel redundancy during the two-stage network training process while enhancing spatial location memory and feature discrimination in the two-dimensional features, we propose a deformable interactive attention module (DIA-Module) to further bolster feature representation capabilities. Additionally, we conduct comparative and transfer experiments on multiple remote sensing datasets, achieving outstanding classification results. Our code is available at: https://github.com/ChenC1027/ADCD-Net.