To excavate adequately the rich information contained in multisource remote sensing data, feature extraction as basic yet important research has two typical applications: one of which is to extract complementary information of multisource data to improve classification; and the other is to extract shared information across sources for domain adaptation. However, typical feature extraction methods require the input represented as vectors or homogeneous tensors and fail to process multisource data represented as heterogeneous tensors. Therefore, the coupled heterogeneous Tucker decomposition (C-HTD) containing two sub-methods, namely coupled factor matrix-based HTD (CFM-HTD) and coupled core tensor-based HTD (CCT-HTD), is proposed to establish a unified feature extraction framework for multisource fusion and domain adaptation. To handle multisource heterogeneous tensors, multiple Tucker models were constructed to extract features of different sources separately. To cope with the supervised and semi-supervised cases, the class-indicator factor matrix was built to enhance the separability of features using known labels and learned labels. To mine the complementarity of paired multisource samples, coupling constraint was imposed on multiple factor matrices to form CFM-HTD to extract multisource information jointly. To extract domain-adapted features, coupling constraint was imposed on multiple core tensors to form CCT-HTD to encourage data from different sources to have the same class centroid. In addition, to reduce the impact of interference samples on domain adaptation, an adaptive sample-weighting matrix was designed to autonomously remove outliers. Using multiresolution multiangle optical and MSTAR datasets, experimental results show that the C-HTD outperforms typical multisource fusion and domain adaptation methods.

Cross-domain object detection in remote sensing suffers from substantial domain gaps arising from differences in resolution, viewing geometry, and imaging modality across sensors and platforms. Existing unsupervised domain adaptive object detection (DAOD) methods typically align source and target features using the detector’s own target-domain representations. However, the extraction of these representations is constrained by the very domain discrepancies they aim to bridge, resulting in noisy and biased features that make alignment unstable. To address this limitation, we propose the Vision-Language model Supervised Domain Adaptor (VLSDA), a domain adaptation framework supervised by a frozen vision-language model (VLM). It leverages a frozen VLM image encoder as an additional and stable semantic domain to guide domain alignment. Our VLM-supervised Prototypical Alignment (VLPA) module stabilizes category-wise alignment through a tri-domain adversarial strategy that jointly aligns source-VLM, target-VLM, and source-target distributions. Complementing this, the Global Cross-domain Contrastive Alignment (GCCA) module enhances intra-class compactness and inter-class separability via supervised contrastive learning. Without requiring any fine-tuning of the VLM, our framework directly mitigates reliance on noisy target features and improves robustness to large distribution shifts. Extensive experiments on multiple cross-domain remote sensing benchmarks demonstrate consistent improvements over state-of-the-art methods, including 68.3% mAP₅₀ on xView→DOTA and 70.5% on HRRSD → SSDD. The code is available at https://github.com/JunhongLu0704/VLSDA.

Semi-supervised object detection (SSOD) in remote sensing is challenged by the accumulation of pseudo-label errors in complex scenes with dense objects and high intra-class variability. While teacher–student frameworks enable learning from unlabeled data, erroneous pseudo-labels such as false positives and missed detections can be reinforced over time, which degrades model performance. To address this issue, we propose the Error-Mitigation Teacher (EMT), a unified framework designed to suppress error propagation during SSOD training. EMT consists of three lightweight modules. First, the Adaptive Pseudo-Label Filtering (APLF) module removes noisy pseudo boxes via a second-stage RCNN and adjusts class-specific thresholds through dynamic confidence filtering. Second, the Confidence-Based Loss Reweighting (CBLR) module reweights training loss by evaluating the teacher model’s ability to reconstruct its own pseudo-labels, using the resulting loss as an indicator of label reliability. Third, the Enhanced Supervised Learning (ESL) module improves class-level balance by adjusting supervised loss weights according to pseudo-label statistics. EMT demonstrates consistent performance gains over representative state-of-the-art SSOD methods on DOTA, DIOR, and SSDD datasets. Notably, EMT achieves a 2.9% absolute mAP₅₀ improvement on DIOR using only 10% of labeled data, without incurring additional inference cost. These results highlight EMT’s effectiveness in improving SSOD for remote sensing.