28 July 2022 in IEEE Transactions on Geoscience and Remote Sensing
Deep learning approaches have contributed to the rapid development of Remote Sensing (RS) image interpretation. The most widely used training paradigm is to utilize ImageNet pre-trained models to process RS data for specified tasks. However, there are issues such as domain gap between natural and RS scenes, and the poor generalization capacity of RS models. It makes sense to develop a foundation model with general RS feature representation. Since a large amount of unlabeled data is available, the self-supervised method has more development significance than the fully supervised method in remote sensing. However, most of the current self-supervised methods use contrastive learning, whose performance is sensitive to data augmentation, additional information, and selection of positive and negative pairs. In this paper, we leverage the benefits of generative self-supervised learning for RS images, and propose a Remote sensing foundation Model framework called RingMo, which consists of two parts. First, a large-scale dataset is constructed by collecting two million RS images from satellite and aerial platforms, covering multiple scenes and objects around the world. Second, we propose an RS foundation model training method designed for dense and small objects in complicated RS scenes. We show that the foundation model trained on our dataset with RingMo method achieves state-of-the-art on eight datasets across four downstream tasks, demonstrating the effectiveness of the proposed framework. Through in-depth exploration, we believe it is time for RS researchers to embrace generative self-supervised learning and leverage its general representation capabilities to speed up the development of RS applications.
31 March 2022 in IEEE Transactions on Geoscience and Remote Sensing
Cross-modal remote sensing text-image retrieval (RSCTIR) has recently become an urgent research hotspot due to its ability of enabling fast and flexible information extraction on remote sensing (RS) images. However, current RSCTIR methods mainly focus on global features of RS images, which leads to the neglect of local features that reflect target relationships and saliency. In this article, we first propose a novel RSCTIR framework based on global and local information (GaLR), and design a multi-level information dynamic fusion (MIDF) module to efficaciously integrate features of different levels. MIDF leverages local information to correct global information, utilizes global information to supplement local information, and uses the dynamic addition of the two to generate prominent visual representation. To alleviate the pressure of the redundant targets on the graph convolution network (GCN) and to improve the model’s attention on salient instances during modeling local features, the denoised representation matrix and the enhanced adjacency matrix (DREA) are devised to assist GCN in producing superior local representations. DREA not only filters out redundant features with high similarity, but also obtains more powerful local features by enhancing the features of prominent objects. Finally, to make full use of the information in the similarity matrix during inference, we come up with a plug-and-play multivariate rerank (MR) algorithm. The algorithm utilizes the $k$ nearest neighbors of the retrieval results to perform a reverse search, and improves the performance by combining multiple components of bidirectional retrieval. Extensive experiments on public datasets strongly demonstrate the state-of-the-art performance of GaLR methods on the RSCTIR task. The code of GaLR method, MR algorithm, and corresponding files have been made available at: https://github.com/xiaoyuan1996/GaLR .
24 November 2022 in IEEE Transactions on Geoscience and Remote Sensing
In recent years, image-level weakly supervised semantic segmentation (WSSS) has developed rapidly in natural scenes due to the easy availability of classification tags. However, limited to complex backgrounds, multi-category scenes, and dense small targets in remote sensing (RS) images, relatively little research has been conducted in this field. To alleviate the impact of the above problems in RS scenes, a self-supervised Siamese network based on an explicit pixel-level constraints framework is proposed, which greatly improves the quality of class activation maps and the positioning accuracy in multi-category RS scenes. Specifically, there are three novel devices in this paper to promote performance to a new level: (a) A pixel-soft classification loss is proposed, which realizes explicit constraints on pixels during the image-level training; (b) A pixel global awareness module, which captures high-level semantic context and low-level pixel spatial information, is constructed to improve the consistency and accuracy of RS object segmentation; (c) A dynamic multi-scale fusion module with a gating mechanism is devised, which enhances feature representation and improves the positioning accuracy of RS objects, particularly on small and dense objects. Experiments on two RS challenge datasets demonstrate that these proposed modules achieve new state-of-the-art results by only using image-level labels, which improve mIoU to 36.79% on iSAID and 45.43% on ISPRS in the WSSS task. To the best of our knowledge, this is the first work to perform image-level WSSS on multi-class RS scenes.
25 April 2022 in IEEE Transactions on Geoscience and Remote Sensing
Despite the extraordinary success of the deep architectures on semantic segmentation for remote sensing (RS) images, they have difficulties in learning new classes from a sequential data stream because of catastrophic forgetting. Continual learning for semantic segmentation (CSS) is an emerging trend for its capability to cope with the above problems effectively. However, old classes from previous steps are collapsed into the background, which further aggravates the challenge of CSS in the RS scene. In this article, we revisit the knowledge distillation (KD) strategy and the characteristics of class-incremental semantic segmentation (CISS) and then present a generalized and effective framework to learn new classes while preserving knowledge of the learned classes. In particular, we propose two novel historical information-guided modules: the feature global perception module and the label reconstruction (LR) module. The former enables the current model to pay more attention to the region related to the old categories identified by the historical information when learning new classes. Meanwhile, the latter retrieves pixels belonging to the learned classes from the background to handle the background shift problem and maintain the high performance of old classes. We have conducted comprehensive experiments on two RS semantic segmentation datasets of Instance Segmentation in Aerial Images Dataset (iSAID) and Gao Fen (GF) challenge semantic segmentation dataset (GCSS). The experimental results outperform the current state-of-the-art methods in most incremental settings, which demonstrates the effectiveness of the proposed framework.
28 November 2023 in IEEE Transactions on Geoscience and Remote Sensing
The rapid development of satellite platforms has yielded copious and diverse multisource data for earth observation, greatly facilitating the growth of multimodal semantic segmentation (MSS) in remote sensing. However, MSS also suffers from numerous challenges: 1) existing inherent defects in each modality due to the different imaging mechanisms; 2) insufficient exploration of the intrinsic characteristics of modalities; and 3) the existence of the huge semantic gap between heterogeneous data causes difficulties in feature fusion. The inability to effectively utilize the rich and diverse information provided by each modality and ignorance of the heterogeneity between modalities will hinder the feature enhancement, and further significantly impacts the semantic segmentation accuracy. Furthermore, neglecting the huge gap makes feature fusion challenging. In this study, we introduce a novel framework for MSS that effectively mitigates the aforementioned problems. Our approach employs a pseudo-Siamese structure for feature extraction. Specifically, we propose a simple yet effective geometric topology structure modeling (GTSM) module to extract geometric relationships and texture information from optical data. Additionally, we present a modality intrinsic noise suppression (MINS) module to fully exploit radiation information and alleviate the effects of unique geometric distortions for synthetic aperture radar (SAR). Furthermore, we present an adaptive multimodal feature fusion (AMFF) module for fully fusing different modality features. Extensive experiments on both WHU-OPT-SAR and DFC23 datasets validate the robustness and effectiveness of the proposed Modality Characteristics-Guided (MoCG) semantic segmentation network compared to other state-of-the-art semantic segmentation methods, including multimodal and single-modal approaches. Our approach achieves the best performance on both datasets, resulting in mean intersection over union (mIoU)/overall accuracy (OA) gains 69.1%/87.5% on WHU-OPT-SAR and 86.7%/97.3% on DFC23.