Rong Liu

Data collected by different modalities can provide a wealth of complementary information, such as hyperspectral image (HSI) to offer rich spectral-spatial properties, synthetic aperture radar (SAR) to provide structural information about the Earth's surface, and light detection and ranging (LiDAR) to cover altitude information about ground elevation. Therefore, a natural idea is to combine multimodal images for refined and accurate land-cover interpretation. Although many efforts have been attempted to achieve multi-source remote sensing image classification, there are still three issues as follows: 1) indiscriminate feature representation without sufficiently considering modal heterogeneity, 2) abundant features and complex computations associated with modeling long-range dependencies, and 3) overfitting phenomenon caused by sparsely labeled samples. To overcome the above barriers, a transformer-based heterogeneously salient graph representation (THSGR) approach is proposed in this paper. First, a multimodal heterogeneous graph encoder is presented to encode distinctively non-Euclidean structural features from heterogeneous data. Then, a self-attention-free multi-convolutional modulator is designed for effective and efficient long-term dependency modeling. Finally, a mean forward strategy is developed in order to avoid overfitting. Based on the above structures, the proposed model is able to break through modal gaps to obtain differentiated graph representation with competitive time cost, even for a small fraction of training samples. Experiments and analyses in three benchmark datasets with various state-of-the-art (SOTA) approaches show the performance of the proposed THSGR. The code will be available in https://github.com/jqyang22.

1. Shallow water bathymetry provides valuable information on underwater depths and potential navigational hazards, which is essential for applications in marine resource development, environmental ...

Accurate detection of individual trees is essential for urban forest management and ecological assessment, yet remains challenging due to the heterogeneous backgrounds, variable sizes of tree crowns, and significant variations across urban scenarios. To address these issues, we propose Tree-SAM, a city-scale individual tree detection architecture built upon the visual foundation model Segment Anything Model (SAM) and equipped with three task-specific modules, i.e., Cross-Correlation Feature Backbone (CCFB), Hierarchical Instance Aggregation Neck (HIAN), and Context-Aware Adaptation Head (CAAH). These modules synergistically fuse general semantics with fine-grained structural cues, enable multi-scale feature aggregation, and adaptively refine predictions based on specific scene contexts. On the GZ-Tree Crown dataset, Tree-SAM achieves F1-scores of 0.762, 0.732, and 0.830, with corresponding AP@50 values of 0.478, 0.454, and 0.526 in the forest, mixed, and urban scenarios, respectively, consistently ranking first across all scenes and demonstrating strong adaptability to diverse intra-city landscapes. Additional evaluations on the BAMFORESTS dataset and the SZ-Dataset further confirm its robustness across varied geographic contexts. Tree-SAM provides a reliable, automated framework for large-scale urban tree mapping, providing reliable data support for urban forest management, carbon stock estimation, and ecological assessment.

Due to the difficulty of obtaining labeled data for hyperspectral images (HSIs), cross-scene classification has emerged as a widely adopted approach in the remote sensing community. It involves training a model using labeled data from a source domain (SD) and unlabeled data from a target domain (TD), followed by inference on the TD. However, variations in the reflectance spectrum of the same object between the SD and the TD, as well as differences in the feature distribution of the same land cover class, pose significant challenges to the performance of cross-scene classification. To address this issue, we propose a dual classification head self-training network (DHSNet). This method aligns class-wise features across domains, ensuring that the trained classifier can accurately classify TD data of different classes. We introduce a dual classification head self-training strategy for the first time in the cross-scene HSI classification field and design a self-training loss based on the prediction of the two classification heads. The proposed approach mitigates the domain gap while preventing the accumulation of incorrect pseudo-labels in the model. Additionally, we incorporate a novel central feature attention mechanism to enhance the model’s capacity to learn scene-invariant features across domains. DHSNet significantly outperforms state-of-the-art methods on three cross-scene HSI datasets, achieving 80.23±1.92% OA on the Houston dataset. The code for DHSNet will be available at https://github.com/liurongwhm.

Satellite derived bathymetry (SDB) has become a cost effective and efficient way to monitor coral reefs on a large scale. Physics-based SDB methods retrieve bathymetry by minimizing a cost function that measures the difference between the observed spectra and radiative transfer model (RTM) simulations. Most existing physics-based SDB methods adopt a single cost function, typically focus on matching spectral amplitude, and ignore other complementary spectral discriminators such as spectral shape. This limitation can compromise retrieval accuracy when spectral amplitude exhibits significant variability due to noise, varying illumination conditions, or benthic endmember variability. To address this problem, this research formulates physics-based SDB as a constrained multi-objective optimization problem (CMOP). The decomposition-based constrained multi-objective differential evolution (DCMODE) algorithm is proposed to simultaneously optimize two cost functions, namely the root-mean-square error (RMSE) and the spectral angle distance (SAD). In DCMODE, evenly distributed weight vectors are produced to guide the optimization under different trade-off preferences. The constrained penalty boundary intersection (CPBI) is introduced as a quantitative criterion for evaluating solutions based on convergence, diversity, and feasibility. Bathymetry results from Landsat-8 OLI images in four study areas in the Xisha Islands indicate that DCMODE achieves higher accuracy than single-objective optimization methods, decreasing the RMSE by 0.03∼0.12 m and the mean absolute error (MAE) by 0.16∼0.19 m. This research demonstrates that multi-objective optimization, as embodied in the DCMODE algorithm, provides a promising framework for enhancing the robustness of physics-based SDB methods. The code and data are publicly available on https://github.com/leic-sysu/RTM-DCMODE.

Deep learning-based hyperspectral target detection (HTD) methods often face the challenge of insufficient prior information and difficulty in distinguishing local and global spectral differences. To address these problems, we propose a self-supervised framework that leverages contrastive learning to reduce dependence on prior knowledge, called the Gated Dual-Path Network with Contrastive Learning (GDPNCL). In this work, we introduce a novel sample augmentation strategy for deep network training, in which each pixel in the scene is processed using a dual concentric window to generate positive and negative samples. In addition, a Gated Dual-Path Network (GDPN) is proposed to effectively extract and discriminate local and global information from the spectra. Moreover, to mitigate the issue of false negative samples within the same class and to enhance the contrast between negative samples, we design a Weight Information Noise contrastive estimation (WIN) loss. The loss leverages the relationship between samples to further help the model learn representations that effectively distinguish targets from diverse backgrounds. Finally, the trained encoder is subsequently employed to extract features from the prior spectrum and test pixels, and the cosine similarity between them serves as the detection metric. Comprehensive experiments on four challenging hyperspectral datasets demonstrate that the GDPNCL outperforms state-of-the-art methods, highlighting its effectiveness and robustness in HTD.

Hyperspectral unmixing aims to resolve the prevalent issue of mixed pixels in hyperspectral imagery and serves as an effective technique for sub-pixel level image interpretation. Recent years have seen the emergence of advanced unmixing algorithms that integrate both spatial and spectral information. However, existing methods mainly focus on spatial context and lack depth in modeling spatial correlations. Both relevant and irrelevant spatial information is introduced into the spectral mixing model for local pixels, with the irrelevant information acting as noise that impacts the unmixing accuracy. To address these challenges, we propose an advanced spectral mixing model, Adaptive Correlation Refined Network (ACR-Net) which integrates refined spatial correlation based on self-attention. A key component of ACR-Net is the Adaptive Correlation Aggregated Decoder (ACAD), which extracts affinity information from the encoder’s feature map and adaptively amplifies the influence of highly correlated regions in the unmixing process. Additionally, the Composite Active Spatial Attention (CASA) Module emphasizes unique spectral characteristics across bands, improving spatial distribution representation and enabling more accurate abundance estimation. We conducted abundant evaluations on six datasets, including three real-world and two synthetic hyperspectral unmixing datasets as well as a large benchmark classification dataset. Extensive experiments have demonstrated that the proposed algorithm can achieve exceptional or comparable unmixing results among various state-of-the-art algorithms. Moreover, the proposed method achieved optimal classification performance on the classic PaviaU dataset, indicating its strong potential for hyperspectral classification task.

Cross-scene classification of hyperspectral images (HSI) aims to classify target domain (TD) data using only labeled source domain (SD) data and unlabeled TD data during training. However, challenges such as spectral shifts across scenes and semantic discrepancies between domains significantly degrade classification performance. To address these issues, domain adaptation (DA) has gained increasing attention in the hyperspectral remote sensing community. This paper proposes a novel framework for cross-scene HSI classification, termed Data Structure-Aware Initialization and Probability-Consistent Self-Training (S2PST) framework. The framework employs batch nuclear-norm maximization to constrain the probability responses of TD outputs, implicitly aligning feature distributions between SD and TD. To enhance the model’s robustness and spectral feature representation ability, we introduce a spectral structure-aware initialization method that integrates the strengths of traditional machine learning and deep learning. Furthermore, to mitigate the model’s bias toward SD training data, we propose a self-supervised training strategy that dynamically incorporates pseudo-labeled TD samples into the training process by comparing the similarity of high-confidence samples in the probability space between SD and TD. Extensive experiments are conducted on the Houston, HyRANK, and Pavia datasets, and compared with several state-of-the-art DA methods. The experiment results demonstrate the effectiveness of the proposed framework. Our code will be available at https://github.com/liurongwhm.