DQDou Quan

Dou Quan

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Dou Quan has a total of 111 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Multi-Expert Learning Framework With the State Space Model for Optical and SAR Image Registration

    by Wei Wang, Dou Quan, Ning Huyan, Chonghua Lv, Shuang Wang, Yunan Li, Licheng Jiao

    Optical and Synthetic Aperture Radar (SAR) image registration is crucial for multi-modal image fusion and applications. However, several challenges limit the performance of existing deep learning-based methods in cross-modal image registration: (i) significant nonlinear radiometric variations between optical and SAR images affect the shared feature learning and matching; (ii) limited textures in images hinder discriminative feature extraction; (iii) the local receptive field of Convolutional Neural Networks (CNNs) restricts the learning of contextual information, while the Transformer can capture long-range global features but with high computational complexity. To address these issues, this paper proposes a multi-expert learning framework with the State Space Model (ME-SSM) for optical and SAR image registration. Firstly, to improve the registration performance with limited textures, ME-SSM constructs a multi-expert learning framework to capture shared features from multi-modal images. Specifically, it extracts features from various transformations of the input image and employs a learnable soft router to dynamically fuse these features, thereby enriching feature representations and improving registration performance. Secondly, ME-SSM introduces a state space model, Mamba, for feature extraction, which employs a multi-directional cross-scanning strategy to efficiently capture global contextual relationships with linear complexity. ME-SSM can expand the receptive field, enhance image registration accuracy, and avoid incurring high computational costs. Additionally, ME-SSM uses a multi-level feature aggregation (MFA) module to enhance the multi-scale feature fusion and interaction. Extensive experiments have demonstrated the effectiveness and advantages of our proposed ME-SSM on optical and SAR image registration. Specifically, ME-SSM improves the correct matching rate (CMR) by 7.14% and 1.95% based on thresholds 1 and 3, respectively, on the SEN1-2 dataset, and increases the CMR by 2.12% based on threshold 3 on the OS dataset. The code is available at https://github.com/Miraitowa515/ME-SSM.


  • Multilevel Feature Adaptation and Aggregation of Vision Foundation Model for Remote Sensing Scene Classification

    by Chonghua Lv, Dan Zhang, Dou Quan, Ning Huyan, Jiaqi Chen, Shuang Wang, Licheng Jiao

    Remote sensing scene classification (RSC) is crucial for image content interpretation and land function identification. Vision Foundation Models (VFMs) have strong general visual representation capabilities, which will enhance the feature extraction from remote sensing images and improve scene classification performances. However, there are obvious discrepancies in datasets and tasks between VFMs and RSC, which will impact the accuracy of scene classification. This paper proposes a VFM-based framework with a multi-level feature adaptation and aggregation approach, labeled as VFM-FAA, for remote sensing scene classification, which can effectively exploit and utilize VFMs’ advantages for the downstream RSC task. Specifically, VFM-FAA adapts the VFM of DINOV2 for RSC from the perspectives of the dataset and task characteristics of the downstream task, respectively. Firstly, considering that there are diverse objects with various sizes in remote sensing images, this paper builds a multi-scale adapter (MSA) based on the frozen VFM, thereby boosting feature extraction from remote sensing images. Secondly, this paper proposes a multi-level feature aggregation module (MFA) to predict the scene category, which can comprehensively capture the discriminative semantic features of various scenes and further improve scene classification accuracy. Experimental results on three public datasets have demonstrated the efficiency and significant advantages of the proposed method on RSC. The proposed method can effectively enhance the generalization of the VFM in the RSC task and reduce the reliance on labeled training data, which has distinct advantages when there is a small amount of labeled training data. Compared with existing methods, VFM-FAA acquires accuracy performance gains by 2.78% in the AID dataset with 20% training samples, and improves the classification accuracy by 1.93% and 2.48% in the NWPU-RESISC45 dataset with 10% and 20% training samples, respectively.

  • C2F-Net: Coarse-to-Fine Feature Alignment for Cross-Channel Automatic Modulation Classification

    by Hantong Xing, Shuang Wang, Chenxu Wang, Dou Quan, Pengtao Li, Huaji Zhou, Licheng Jiao

    Despite the impressive performance of deep learning (DL)-based automatic modulation classification (AMC) models, their effectiveness decrease significantly when confronted with the complex channel conditions within non-cooperative communications. In response, researchers have proposed a series of adversarial training-based domain adaptation (DA) methods to align the feature distributions of different channels. However, the inherently unstable training process of adversarial training hinders these models from achieving satisfactory results, particularly when there are significant changes in the application scenarios and channel conditions of communication devices. To overcome this challenge, we introduce C2F-Net, a coarse-to-fine feature alignment model for cross-channel AMC. During coarse alignment, C2F-Net is expected to use masking operations to force the model to learn intra-domain features and inter-domain alignment relationships from incomplete signals, thereby improving the channel invariance of the feature space. Benefit from this, during fine alignment, the self-training-based DA method is capable of generating target domain pseudo-labels with enhanced confidence. These pseudo-labels serve as categorical supervision information for feature learning, enabling the construction of a more accurate and refined feature space, which facilitates more precise feature alignment. Comprehensive experiments conducted across five datasets demonstrate that C2F-Net achieves an average performance improvement of up to 8.99% compared to baseline methods.


  • Generalization-Aware Remote Sensing Change Detection via Domain-Agnostic Learning

    by Qi Zang, Shuang Wang, Dong Zhao, Dou Quan, Yang Hu, Zhi Yang, Licheng Jiao

    Change detection has essential significance for the region's development, in which pseudo-changes between bitemporal images induced by imaging environmental factors are key challenges. Existing transformation-based methods regard pseudo-changes as a kind of style shift and alleviate it by transforming bitemporal images into the same style using generative adversarial networks (GANs). However, their efforts are limited by two drawbacks: 1) Transformed images suffer from distortion that reduces feature discrimination. 2) Alignment hampers the model from learning domain-agnostic representations that degrades performance on scenes with domain shifts from the training data. Therefore, oriented from pseudo-changes caused by style differences, we present a generalizable domain-agnostic difference learning network (DonaNet). For the drawback 1), we argue for local-level statistics as style proxies to assist against domain shifts. For the drawback 2), DonaNet learns domain-agnostic representations by removing domain-specific style of encoded features and highlighting the class characteristics of objects. In the removal, we propose a domain difference removal module to reduce feature variance while preserving discriminative properties and propose its enhanced version to provide possibilities for eliminating more style by decorrelating the correlation between features. In the highlighting, we propose a cross-temporal generalization learning strategy to imitate latent domain shifts, thus enabling the model to extract feature representations more robust to shifts actively. Extensive experiments conducted on three public datasets demonstrate that DonaNet outperforms existing state-of-the-art methods with a smaller model size and is more robust to domain shift.


  • Disentangling Domain-Invariant Features From Few Samples for Automatic Modulation Classification

    by Hantong Xing, Shuang Wang, Chenxu Wang, Dou Quan, Yi Xu, Huaji Zhou, Hua Xu, Licheng Jiao

    The existing deep learning (DL)-based automatic modulation classification (AMC) models are difficult to apply in non-cooperative communication systems, primarily due to two major challenges: 1) domain shift: the variable usage scenarios of frequency devices often result in data distribution shifts during signal transmission, significantly compromising the performance of existing models and 2) sample scarcity: collecting large amounts of samples to support the training of DL models is impractical as it may involve issues of legal privacy and resource consumption. Existing works has addressed these two challenges as separate tasks, making it difficult to implement a comprehensive solution. Consequently, this paper proposes a unified framework named signal feature disentangling network (SFDN). By leveraging dual constraints at the instance and feature levels, SFDN effectively disentangles signal features into domain-specific and content-related components. The content features are subsequently utilized for classification tasks, while the domain features are disregarded, thereby mitigating the challenge of domain shift. Furthermore, the exchange and recombination of domain and content features facilitate the generation of cross-domain samples. These synthetically created samples are then employed in a supervised manner, significantly alleviating the challenge of sample scarcity. The comprehensive experimental verification under three typical channel conditions demonstrates the progressiveness of our SFDN.


  • AFLNet: Auxiliary Feature Learning-Guided Cross-Channel Automatic Modulation Classification

    by Hantong Xing, Shuang Wang, Chenxu Wang, Dou Quan, Hanlin Mo, Luyang Mei, Huaji Zhou, Licheng Jiao

    This paper conducted a thorough investigation into the primary difficulty of the cross-channel automatic modulation classification (AMC) task by examining data distribution and feature space of different channel conditions. We concluded that the disruption of the target channel feature space structure breakdown the mapping relationship across channels, serving as the main contributor to model performance degradation. Based on the above conclusion, in order to improve the performance of cross-channel AMC, we introduce the Auxiliary Feature Learning-Guided Network (AFLNet). This network improves the structure of the target feature space through two uniquely designed tasks and facilitates efficient cross-domain alignment via a collaborative alignment mechanism. Specifically, AFLNet integrates similarity-based and confidence-based auxiliary feature learning tasks to enhance the discriminability of the target feature space and maintain the correspondence of category structures across different channels, thereby reducing the difficulty of feature alignment. The collaborative alignment mechanism combines adversarial training-based and self-training-based feature alignment methods, leveraging their mutually reinforcing effect and complementary strengths in global alignment and class-level alignment to enhance overall alignment performance. We carried out extensive experiments across four scenarios characterized by substantial channel variations, verifying that AFLNet achieves state-of-the-art with accuracy improvement of up to 9.71%.


  • Burden-Free Distillation From Foundation Model for Efficient Remote Sensing Change Detection

    by Shuang Wang, Chonghua Lv, Dou Quan, Ning Huyan, Xianwei Cao, Jingxi Sun, Licheng Jiao

    Applying vision foundation models to remote sensing change detection (CD) has attracted extensive research attention. These studies employ inherent general knowledge from vision foundation models to enhance CD performance. Existing methods explicitly employ the foundation model as a feature extractor while designing additional learnable modules to bridge the task gap. However, these methods substantially increase the computational burden and memory demand in the inference. This paper therefore focuses on addressing the core challenge of effectively leveraging the knowledge from vision foundation models to enhance CD performance while maintaining computational efficiency. Instead of explicitly utilizing the foundation model, we propose Burden-Free Distillation (BFD), an architecture-agnostic foundation model-based distillation framework for efficient CD. BFD transfers the general knowledge from foundation models to task-specific models, thereby eliminating the dependency on foundation models during inference. Specifically, BFD transfers the foundation model knowledge through Dual-temporal Feature Matching module (DFM). This module enables multi-level feature alignment by computing pixel-wise spatial similarity between the foundation models’ general features and the CD models’ task-specific features. Additionally, we leverage patch contrastive distillation, which transfers localized structural patterns to the CD model to further mitigate task discrepancies between foundation models and CD models. We conduct extensive experiments across multiple foundation models and CD architectures, experimental results demonstrate that BFD effectively adapts the knowledge of foundation models to CD tasks without additional computational burden. Compared to other foundation model-based CD methods, BFD reduces the model parameters by 80.3% and improves IoU by 1.78% on the S2Looking dataset. The code is available at https://github.com/Younger-hua/Burden-Free-Distillation.