Fine-grained ship classification plays an important part in many military and civilian applications. However, it is often costly to obtain images of ships, making it difficult to procure large numbers of such images. This difficulty poses challenges to machine learning procedures that require ship images. Commonly, only a few input images are available for certain types of ships, which leads to the poor generalization of trained models. Therefore, few-shot fine-grained ship classification is an important (but significantly challenging) task in machine learning. In this study, we propose a novel foreground-aware feature map reconstruction network that is simple, effective, and scalable. We reconstruct the query features from support features using ridge regression and predict the distribution of the categories of query images between the reconstructed and real query features by comparing the weighted distances with foreground weights. The foreground weights indicate the percentages of foreground information in the feature map locations. We propose two methods for calculating the foreground weights: a non-parametric method and a parametric method. Our proposed network achieves state-of-the-art results on both the fine-grained ship classification dataset FGSCR and the natural fine-grained bird classification dataset CUB.

In complex environments, traditional multi-target tracking methods often encounter challenges such as strong clutter interference and interruptions in target trajectories, which can result in insufficient tracking accuracy and robustness. To address these issues, this paper presents an improved multi-target tracking algorithm, termed Q-IMM-MHT. This method integrates Multiple Hypothesis Tracking (MHT) with Interactive Multiple Model (IMM) and introduces a Q-learning-based adaptive model switching strategy to dynamically adjust model selection in response to variations in the target’s motion patterns. Furthermore, the algorithm utilizes Support Vector Machines (SVMs) for anomaly detection and trajectory recovery, thereby enhancing the accuracy of data association and the overall robustness of the system. Experimental results indicate that under high noise conditions, the Root Mean Square Error (RMSE) of position estimation decreases to 0.74 pixels, while the RMSE of velocity estimation falls to 0.04 pixels/frame. Compared to traditional methods such as the Unscented Kalman Filter (UKF), IMM, and CIMM, the RMSE is reduced by at least 10.84% and 42.86%, respectively. In scenarios characterized by target trajectory interruptions and clutter interference, the algorithm maintains an association accuracy exceeding 46.3% even after 30 frames of interruption, significantly outperforming other methods. These findings demonstrate that the Q-IMM-MHT algorithm offers substantial performance improvements in multi-target tracking tasks within complex environments, effectively enhancing both tracking accuracy and stability, with considerable application value and extensive potential for future use.

The rapid development of remote sensing technology has brought about a sharp increase in the amount of remote sensing image data. However, due to the satellite’s limited hardware resources, space, and power consumption constraints, it is difficult to process massive remote sensing images efficiently and robustly using the traditional remote sensing image processing methods. Additionally, the task of satellite-to-ground target detection has higher requirements for speed and accuracy under the conditions of more and more remote sensing data. To solve these problems, this paper proposes an extremely efficient and reliable acceleration architecture for forward inference of the YOLOX-s detection network an on-orbit FPGA. Considering the limited onboard resources, the design strategy of the parallel loop unrolling of the input channels and output channels is adopted to build the largest DSP computing array to ensure a reliable and full utilization of the limited computing resources, thus reducing the inference delay of the entire network. Meanwhile, a three-path cache queue and a small-scale cascaded pooling array are designed, which maximize the reuse of on-chip cache data, effectively reduce the bandwidth bottleneck of the external memory, and ensure an efficient computing of the entire computing array. The experimental results show that at the 200 MHz operating frequency of the VC709, the overall inference performance of the FPGA acceleration can reach 399.62 GOPS, the peak performance can reach 408.4 GOPS, and the overall computing efficiency of the DSP array can reach 97.56%. Compared with the previous work, our architecture design further improves the computing efficiency under limited hardware resources.

Interpreting airborne remote sensing images plays an important role in aviation control and battlefield situational awareness. However, highly dynamic aircraft detection remains challenging, owing to variable object sizes, flexible attitudes, and motion blur. This study develops a multi-scale airborne aircraft dataset benchmark to overcome aircraft detection challenges, such as high intraclass variance, multiple scales and angles, motion blur, and partial occlusion. We also propose a trained-from-scratch aircraft detector, the local feature correlation single shot multibox detector (LFC-SSD), to detect multi-scale aircraft. The LFC-SSD comprises a local correlation feature extraction module, called “Right-Residual,” and a feature fusion module using a reverse feature pyramid network (R-FPN). Right-Residual extends the global receptive field by aggregating contextual information while learning non-adjacent region features efficiently. R-FPN utilizes multi-path information transfer horizontally with recursive integration to enhance the robust representation of the location information of multi-scale object features. In addition, a specific default boxes method is designed for remote-sensing images of aircraft. Extensive experimental results confirm the significant improvement of the proposed method over several existing state-of-the-art methods.

Path planning for stratospheric airships in dynamic wind fields is challenging due to complex wind variations and strict nighttime energy constraints. This paper proposes a hybrid Level-Set Particle Swarm Optimization (LS-PSO) framework. Firstly, it employs PSO to search iteratively for a propulsion velocity sequence in the velocity domain, with a multi-objective fitness function that integrates reachability, energy consumption and time cost to evaluate each velocity sequence. Then, the reachability of each candidate sequence is numerically solved by the Level Set forward evolution. To improve optimization efficiency, we proposed a multi-resolution grid adaptive strategy for forward evolutions. Finally, with the optimal velocity sequence, the optimal path is generated once by the Level Set backtrack processing. To validate the resulting methodology, we used a benchmark case of a dynamic complex four-gyre flow, described by mathematical formulas, with the optimal day–night path identified by GPOPS-II. The results show the LS-PSO solution has comparable accuracy, with a trajectory deviation less than 3%. Then, we tested the methodology in the stratospheric wind flows using ERA5 reanalysis data. The results demonstrate that our path planning methodology provides a computationally efficient and optimal energy–time solution for autonomous stratospheric airships, while conforming to reachability and strict nighttime energy constraints.

This paper seeks to address the limitations of conventional remote sensing data dissemination algorithms, particularly their inability to model fine-grained multi-modal heterogeneous feature correlations and adapt to dynamic network topologies under resource constraints. This paper proposes multi-modal-MAPPO, a novel multi-modal deep reinforcement learning (MDRL) framework designed for a proactive data push in large-scale integrated satellite–terrestrial networks (ISTNs). By integrating satellite cache states, user cache states, and multi-modal data attributes (including imagery, metadata, and temporal request patterns) into a unified Markov decision process (MDP), our approach pioneers the application of the multi-actor-attention-critic with parameter sharing (MAPPO) algorithm to ISTNs push tasks. Central to this framework is a dual-branch actor network architecture that dynamically fuses heterogeneous modalities: a lightweight MobileNet-v3-small backbone extracts semantic features from remote sensing imagery, while parallel branches—a multi-layer perceptron (MLP) for static attributes (e.g., payload specifications, geolocation tags) and a long short-term memory (LSTM) network for temporal user cache patterns—jointly model contextual and historical dependencies. A dynamically weighted attention mechanism further adapts modality-specific contributions to enhance cross-modal correlation modeling in complex, time-varying scenarios. To mitigate the curse of dimensionality in high-dimensional action spaces, we introduce a multi-dimensional discretization strategy that decomposes decisions into hierarchical sub-policies, balancing computational efficiency and decision granularity. Comprehensive experiments against state-of-the-art baselines (MAPPO, MAAC) demonstrate that multi-modal-MAPPO reduces the average content delivery latency by 53.55% and 29.55%, respectively, while improving push hit rates by 0.1718 and 0.4248. These results establish the framework as a scalable and adaptive solution for real-time intelligent data services in next-generation ISTNs, addressing critical challenges in resource-constrained, dynamic satellite–terrestrial environments.