The detection and localization of craters on the Moon and other planets play an essential role in planet landing, spacecraft navigation, and geologic study. Historically, craters detection involves manually measuring the size and placement of craters in surface images. In this paper, we propose an automated pipeline named CraterNet to detect craters on the Moon and Mercury from DEM images. Firstly, the convolutional neural network (CNN) based object detection model is trained and tested to detect the craters in a supervised manner. To address the domain discrepancy between the source and target data, an unsupervised domain adaptation (UDA) approach combined with domain randomization is suggested. A causal inference-based feature matching (CIFM) approach integrated with histogram matching is then developed to improve the effectiveness the unsupervised crater detection. The DeepMoon crater dataset and the unsupervised Mercury crater DEMs are introduced in this paper to illustrate the applicability and efficacy of the designed method. Results indicate that (1) The developed approach demonstrates a high performance of supervised crater detection on the DeepMoon dataset, with the F1 and AP scores of 0.786 and 0.804, respectively. (2) The detection model is transferred and performs well on the Mercury dataset in which craters are of different sizes and shapes, as the Precision, Recall, and F1 score for ellipse-shape craters are 0.734, 0.773, and 0.753, respectively. (3) The proposed CraterNet outperforms other deep learning-based segmentation and detection models in terms of crater detection and localization scores.

Video satellites have recently become an attractive method of Earth observation, providing consecutive images of the Earth’s surface for continuous monitoring of specific events. The development of on-board optical and communication systems has enabled the various applications of satellite image sequences. However, satellite video-based target tracking is a challenging research topic in remote sensing due to its relatively low spatial and temporal resolution. Thus, this survey systematically investigates current satellite video-based tracking approaches and benchmark datasets, focusing on five typical tracking applications: traffic target tracking, ship tracking, typhoon tracking, fire tracking, and ice motion tracking. The essential aspects of each tracking target are summarized, such as the tracking architecture, the fundamental characteristics, primary motivations, and contributions. Furthermore, popular visual tracking benchmarks and their respective properties are discussed. Finally, a revised multi-level dataset based on WPAFB videos is generated and quantitatively evaluated for future development in the satellite video-based tracking area. In addition, 54.3% of the tracklets with lower Difficulty Score (DS) are selected and renamed as the Easy group, while 27.2% and 18.5% of the tracklets are grouped into the Medium-DS group and the Hard-DS group, respectively.

Deep learning-based approaches for land cover segmentation rely on supervision with pixel-level ground truth, but may not generalize well to unseen image domains. Due to the tedious and labor-intensive labeling process, transferring the models trained with label-rich source data to nonannotated target data becomes a popular problem in recent years. Because of the domain shift, the difference between the source and target distributions can degrade the accuracy on target data if the training occurs directly in a source domain without proper domain adaptation (DA). In this letter, we propose a U-Net based network for DA in the context of semantic segmentation. The model is trained in the source domain with ground truth and test in the target domain without any annotations. We introduce the layer alignment method and the feature covariance loss function to alleviate the domain shift between different domains. To further enhance the adapted model, we adopt a self-training method to improve segmentation performance. Experimental results on the images from the 2018 IEEE Geoscience and Remote Sensing Society (GRSS) data fusion contest and the International Society for Photogrammetry and Remote Sensing (ISPRS) 2-D semantic labeling contest data set reveal the effectiveness of the proposed model. By reducing the domain distribution difference, our method shows better performance compared with the mainstream unsupervised DA methods.

Small-scale low-altitude unmanned aerial vehicles (UAVs) equipped with perception capability for military targets will become increasingly essential for strategic reconnaissance and stationary patrols in the future. To respond to challenges such as complex terrain and weather variations, as well as the deception and camouflage of military targets, this paper proposes a hybrid detection model that combines Convolutional Neural Network (CNN) and Transformer architecture in a decoupled manner. The proposed detector consists of the C-branch and the T-branch. In the C-branch, Multi-gradient Path Network (MgpNet) is introduced, inspired by the multi-gradient flow strategy, excelling in capturing the local feature information of an image. In the T-branch, RPFormer, a Region–Pixel two-stage attention mechanism, is proposed to aggregate the global feature information of the whole image. A feature fusion strategy is proposed to merge the feature layers of the two branches, further improving the detection accuracy. Furthermore, to better simulate real UAVs’ reconnaissance environments, we construct a dataset of military targets in complex environments captured from an oblique perspective to evaluate the proposed detector. In ablation experiments, different fusion methods are validated, and the results demonstrate the effectiveness of the proposed fusion strategy. In comparative experiments, the proposed detector outperforms most advanced general detectors.

In order to enhance the efficiency of the image transmission system and the robustness of the optical imaging system of the Association of Sino-Russian Technical Universities satellite, a new framework of on-board cloud detection by utilizing a lightweight U-Net and JPEG compression strategy is described. In this method, a careful compression strategy is introduced and evaluated to acquire a balanced result between the efficiency and power consuming. A deep-learning network combined with lightweight U-Net and Mobilenet is trained and verified with a public Landsat-8 data set Spatial Procedures for Automated Removal of Cloud and Shadow. Experiment results indicate that by utilizing image-compression strategy and depthwise separable convolutions, the maximum memory cost and inference speed are dramatically reduced into 0.7133 Mb and 0.0378 s per million pixels while the overall accuracy achieves around 93.1%. A good possibility of the on-board cloud detection based on deep learning is explored by the proposed method.

Due to its highly complementary information about remote sensing, synthetic aperture radar (SAR) and optical imagery matching have drawn much attention in recent years. Compared with traditional methods, deep learning-based SAR-optical image matching models largely rely on supervision with ground truths, where the matching accuracy suffers because of unseen image domains. To mitigate loads in burdensome labeling tasks, transferring deep learning models trained with annotated source domains to nonannotated target domains has attracted great concern. Due to the domain gap, the difference between the source and target domains is likely to deteriorate the matching accuracy on target data if the training process is directly conducted without proper domain adaptation (DA). In this research, a Siamese DA (SDA) approach with a combined loss function is developed in the context of multimodality image matching. Then, a novel rotation/scale-invariant transformation module with regression modules is designed to extract rotation/scale-equivariant features. Finally, the causal inference-based self-learning method and the multiresolution histogram matching approach are employed to enhance the unsupervised matching performance. Experimental results on the RadarSat/Planet dataset and the Sentinel-1/2 dataset demonstrate that the developed model can achieve competitive matching performance with a low overlap ratio between domains and little data labeling. By alleviating the domain discrepancy, the developed model drastically reduces the average L2 score of the unsupervised matching from 9.576 to 0.658, while the less-than-one-pixel matching error rate is enhanced from 66.3% to 90.6%.