Song Dai

Dr. Song Dai


Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 October 2024 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
DSDongmei Song
Dongmei Song
FQFuhou Qin
Fuhou Qin
BWBin Wang
Bin Wang
SDSong Dai
Song Dai
Due to the high operational complexity of manual sample labeling for hyperspectral images (HSI), few-shot learning (FSL) has been introduced to cope with the lack of training samples in the field of HSI classification and achieved good results by virtue of its excellent performance. However, factors such as category labeling noise, category distinguishability in the metric space, and completeness of effective feature mining are still the primary considerations that significantly affect the stability and robustness of FSL. Therefore, this study proposes a framework of Few-Shot Learning with Label smoothing and Metric space optimization (LMFSL) for HSI classification. The framework first incorporates a category label smoothing strategy into FSL, which mitigates the effect of noise by constructing a novel category label smoothing module (CLSM), to reduce the confidence of the classifier. Meanwhile, the study also designs a metric space optimization module (MSOM), which prompts similar samples in the metric space to largely aggregate together by maximizing the intra-class similarity and minimizing the inter-class similarity in a more flexible way, so as to improve the decision boundary of the model and enhance the recognition performance of the model. Furthermore, to achieve effective feature extraction in the context of few labeled samples, a lightweight feature extractor, LFE-MAs, incorporating multiple attention mechanisms is designed to extract the HSI features with high efficiency and low computational cost. Experimental results on four public HSI datasets show that LMFSL outperforms other state-of-the-art methods in terms of classification accuracy with limited labeled samples and has lower computational complexity.

25 August 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Verified
Journal Article
Open access
DSDongmei Song
Dongmei Song
YWYang Wang
Yang Wang
RXRonghan Xu
Ronghan Xu
BWBin Wang
Bin Wang
SDSong Dai
Song Dai
YBYuhai Bao
Yuhai Bao
As an emerging remote sensing technology, Global Navigation Satellite System Reflectometry (GNSS-R) offers moderate penetration capabilities and supports all-weather observations. It has been widely applied in soil moisture retrieval. However, GNSS-R-based soil moisture inversion faces significant challenges due to the marked heterogeneity, complex feature associations, and difficulty in unified modeling among Delay-Doppler Map (DDM) imagery, time delay waveforms, and auxiliary features. To address these issues, this study introduces a multi-modal fusion strategy combined with a deep learning approach that possesses strong nonlinear modeling capabilities, and proposes a novel GNSS-R soil moisture retrieval model termed the Multi-modal Cross-Branch Fusion Network (MCBF-Net). MCBF-Net consists of three parallel branches designed to process different modalities of input data: a one-dimensional convolutional neural network (1D CNN) module for efficiently extracting features from time delay waveforms; a multi-scale fusion Transformer module for capturing the global features of DDM images; and a backpropagation (BP) neural network module that leverages its powerful nonlinear mapping capabilities to deeply learn complex patterns in the auxiliary data. Additionally, a multi-modal feature fusion module and a fusion loss function are embedded within the tri-branch architecture to ensure effective integration of features across modalities and to enhance the synergy among the network branches. Through this design, the model enables deep and comprehensive feature extraction and fusion from time delay waveforms, DDM imagery, and auxiliary information, significantly improving the accuracy of soil moisture retrieval. This study uses the Yellow River Delta region as the experimental area, focusing on the urgent needs of ecological environmental protection and sustainable agricultural development in the region. The soil moisture observation data provided by the Soil Moisture Active Passive (SMAP) satellite were used to validate the effectiveness of the proposed method. The retrieved results achieved a root mean square error (RMSE) of 0.0284 cm³/cm³ and a correlation coefficient (R) of 0.9232, indicating high accuracy and reliability. Furthermore, the study analyzed the spatiotemporal distribution characteristics of soil moisture in the Yellow River Delta, offering a feasible and efficient solution for regional-scale soil moisture inversion and refined monitoring.

22 May 2025 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
SDSong Dai
Song Dai
DSDongmei Song
Dongmei Song
BWBin Wang
Bin Wang
CWChen Weimin
Chen Weimin
Multisource data fusion offers great potential for land cover classification. However, the substantial differences in data structures and content representations across various remote sensing sources present significant challenges in heterogeneous feature extraction and information fusion, ultimately constraining the effectiveness of fusion-based classification. To address the aforementioned limitations, this article proposes a multisource data fusion classification method based on a cross-modal cascaded encoder-decoder network (CCEnd-Net). The proposed algorithm comprises two primary components: a blur feature extraction module and a cascaded encoder-decoder feature fusion module. Specifically, the blur feature extraction module utilizes multidimensional convolution to extract deep features from multisource data and incorporates a blur pooling module to enhance aliasing resistance. This approach mitigates the original spatial discrepancies among heterogeneous data while preventing feature distortion. Meanwhile, the cascaded encoder-decoder feature fusion module reconstructs multisource data features by integrating multiscale channel-spatial interaction attention (MCIA)-enhanced convolutional neural networks (CNNs) with a transformer-based cross-modal fusion (TCMF) block. Additionally, a multilevel fusion strategy is employed to comprehensively exploit the complementary information from different remote sensing data sources, thereby improving classification performance. To validate the effectiveness of the proposed method, comprehensive experiments were conducted on four benchmark datasets covering light detection and ranging (LiDAR), hyperspectral imaging (HSI), synthetic aperture radar (SAR), and very high resolution (VHR) data. The proposed CCEnd-Net was systematically evaluated against state-of-the-art models, including transformer-based architectures, CNNs, and traditional classifiers. The experimental results demonstrate that CCEnd-Net achieves superior performance across all datasets.

18 April 2025 in IEEE Sensors Journal
Verified
Journal Article
Closed access
JZJingyao Zhang
Jingyao Zhang
XCXuerong Cui
Xuerong Cui
JLJuan Li
Juan Li
BJBin Jiang
Bin Jiang
LLLei Li
Lei Li
SDSong Dai
Song Dai
Side-scan sonar (SSS) is a commonly used sensor device in underwater pipeline detection tasks. In the processing of SSS waterfall images for underwater pipeline detection, uneven grayscale often occurs due to factors such as sonar energy attenuation and underwater channel interference. Existing grayscale correction techniques struggle to meet the specific needs of underwater pipeline detection, as they fail to fully utilize pipeline characteristic information and have limitations in adapting to the complex underwater environment. To address these challenges, this paper proposes a dual-branch adaptive correction (DBAC) algorithm for adaptive grayscale correction of SSS images containing underwater pipelines. Unlike existing methods, this approach decomposes SSS images into low-frequency and high-frequency components using Shearlet transform. A pipeline-aware adaptive gamma correction (PAGC) algorithm and an adaptive dictionary update sparse dictionary learning (ADASDL) algorithm are designed to process these components accordingly. The PAGC algorithm uses pipeline contour and large-scale texture features in the low-frequency components as specific constraints to dynamically adjust the gamma value, enhancing the grayscale correction effect. Meanwhile, the noise in the high-frequency components is effectively reduced using the ADASDL algorithm. This innovative dual-branch strategy improves the adaptability of grayscale correction methods in complex underwater environments while making full use of pipeline feature information. Experimental results show that the proposed method outperforms four traditional algorithms and four newly proposed algorithms in terms of visual quality across six datasets, validating its effectiveness and innovation in grayscale correction for underwater pipeline side-scan sonar images.

14 January 2026 in Marine Pollution Bulletin
Verified
Journal Article
Closed access
JZJingyao Zhang
Jingyao Zhang
XCXuerong Cui
Xuerong Cui
SDSong Dai
Song Dai
JLJuan Li
Juan Li
LLLei Li
Lei Li

24 November 2025 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
SDSong Dai
Song Dai
DSDongmei Song
Dongmei Song
ESErdenesukh Sumiya
Erdenesukh Sumiya
RXRonghan Xu
Ronghan Xu
MYMei Yong
Mei Yong
YBYuhai Bao
Yuhai Bao
Global navigation satellite system reflectometry (GNSS-R) is widely used for soil moisture retrieval. To mitigate heterogeneity between delay–doppler map (DDM) and auxiliary features and to impose physical constraints in fusion modeling, we propose a physics-guided cross-feature fusion network (PCF-Net) for the Yellow River Delta, comprising a multi-feature input, cross-feature fusion, and a physics-constrained retrieval module. Specifically, the multi-feature input module employs convolutional neural networks (CNNs) to extract spatial–local features from DDM and deep semantic information from auxiliary data, enabling structural alignment and unified embedding of heterogeneous modalities. The cross-feature fusion module adopts mamba-based cross-feature fusion (MCF) block and transformer-based cross-feature fusion (TCF) block to achieve bidirectional interaction and deep coupling between one-dimensional auxiliary features and DDM image features. The physics-constrained retrieval module constructs a physics-consistency loss between measured and simulated GNSS-R surface reflectivity to regularize the training process. We further construct a multi-source GNSS-R dataset from cyclone global navigation satellite system (CYGNSS), Tianmu-1, and fengyun-3E (FY-3E) and validate against soil moisture active passive (SMAP). The joint use of multi-source GNSS-R data markedly improves retrieval accuracy, achieving R of 0.9805 and RMSE of 0.0176 cm³/cm³, demonstrating effectiveness at the regional scale.