The widespread deployment of low-altitude drones in Internet of Things (IoT) applications, such as smart transportation and urban security, demands efficient, real-time visual perception on resource-constrained edge devices. A critical challenge in this domain is the precise detection and localization of small objects against complex backgrounds. Low-altitude UAV imagery is particularly challenging due to blurred small-target edges, weak anti-interference for medium and large targets, and fragmented multi-scale features. Existing convolutional and general-purpose Transformer-based detectors exhibit significant limitations in localization accuracy and robustness under these conditions. To address these issues, we propose E²-Former, a novel object detection model based on the Detection Transformer (DETR) framework, designed for real-time performance on computational-edge devices. Our model integrates three core components: an Edge-Enhanced Backbone that augments sensitivity to contours while preserving deep semantics; a Polarized Dynamic Multi-Feature Fusion (PDMF) Transformer that leverages dual-path polarized attention and frequency-domain modulation to enhance local-global feature modeling and suppress background noise; and an Edge-aware Path Aggregation Network (E-PAN) that uses bidirectional gating and multi-level context interaction to resolve feature fragmentation and promote fine-grained, cross-scale integration. On three challenging low-altitude datasets—VisDrone2019, UAVDT, and CODrone—E²- Former achieves leading AP₅₀ scores of 48.9%, 41.2%, and 33.0%, respectively, outperforming existing mainstream methods in both detection accuracy and robustness. Practical deployment and scene detection experiments further validate its impressive performance in real-world scenarios, establishing a new paradigm for building efficient and reliable low-altitude drone perception systems.

Small object detection in unmanned aerial vehicle (UAV) remote sensing images remains challenging due to large-scale variations, dense object distributions, and complex background interference. Although Transformer-based detectors have improved global context modeling for remote sensing object detection, many existing designs still rely on spatial geometric relationships and conventional cross-level fusion, which may limit the aggregation of spatially scattered small-object features and introduce background interference. To address these issues, this paper proposes a Multi-granularity Prompt and Intensity Guidance Detection Transformer (MPI-DETR), an efficient end-to-end Transformer-based detector for small-object detection in UAV remote sensing images. MPI-DETR consists of three key components: a Dual-Stream Ranked Self-Attention (DRSA) module for intensity-ordered global feature aggregation, a Bilateral Tanh Gating and Cosine Attention Feature Alignment Module (BTC-FAM) for noise-resistant cross-level alignment, and a Prompt-driven Multi-granularity Fusion (PMGF) module for enhancing weak small-object details. Experiments on AI-TOD, DIOR, and NWPU VHR-10 demonstrate that MPI-DETR achieves $A{P}_{50}$ scores of $43.8\%$, $87.5\%$, and $92.5\%$, respectively. Compared with the RT-DETR-R18 baseline, MPI-DETR improves $A{P}_{50}$ by 3.0, 1.0, and 3.9 percentage points on the three datasets, respectively, and increases $A{P}_{\mathrm{S}}$ by 2.9, 2.9, and 11.6 percentage points. It also surpasses the strongest compared models by 1.7, 1.0, and 1.3 percentage points in $A{P}_{50}$ on AI-TOD, DIOR, and NWPU VHR-10, respectively. These results indicate that MPI-DETR provides a robust and efficient solution for small-object perception in complex UAV remote sensing applications, especially in scenes with dense targets, background interference, and limited computational resources.

Pansharpening plays an important role in remote sensing image processing. Its purpose is to fuse a high-spatial-resolution panchromatic (PAN) image and a low-spatial-resolution multispectral (LRMS) image, thereby reconstructing a high-resolution multispectral (HRMS) image with both high spatial clarity and high spectral fidelity. In recent years, diffusion models have shown great potential in image generation. However, existing diffusion-based pansharpening methods usually adopt a fixed denoising strategy, making it difficult to adapt to the stage-wise changes in the denoising process and complex degradation distributions. Based on this, we propose PanDiM, an efficient generative framework for pansharpening. Specifically, we reformulate pansharpening as a high-frequency residual restoration process constrained by multimodal conditions. To improve the response accuracy of the model in complex regions, we design a Degradation-Posterior Guidance Module (DPGM), which extracts dual-scale physical detail priors from the PAN image, explicitly infers the degradation posterior, and converts it into dynamic control variables to adaptively regulate the state evolution of Mamba. In addition, we propose a time-aware mechanism, which allows temporal information to directly intervene in posterior estimation and state-space modeling, so as to accurately match the modeling requirements of different denoising stages. Considering the characteristics of residual reconstruction, we further propose a frequency-decoupled loss (FDL), which separates low- and high-frequency components in the frequency domain and applies targeted constraints. This significantly enhances the model’s ability to represent textures and achieves more robust spectral fidelity. Extensive experiments on three benchmark datasets, including WorldView-3, GaoFen-2, and QuickBird, show that PanDiM significantly outperforms existing mainstream methods in both reduced-resolution and full-resolution evaluations, providing a new solution for high-fidelity pansharpening in complex scenarios.

Remote sensing images often exhibit characteristics such as significant scale variation, dense distribution, and complex background textures, which lead to the obscuring of small target features and result in missed and false detections in occlusion scenarios. To address this, we propose a lightweight remote sensing dense small target detection network, HSRNet. This method introduces a Sparse Adaptive Attention Enhancement Module (SASR Transformer) in the deep features to strengthen effective context and suppress background interference. An Asymmetric Feature Aggregation Module (AFAM) is designed at the neck stage to improve multi-scale feature aggregation quality. Additionally, the detection head incorporates hypergraph relation modeling to construct higher-order semantic topology and better distinguish adjacent dense targets through gated distribution. Experimental results on the DOTA1.0, DIOR, and NWPU VHR-10 datasets demonstrate that HSRNet achieves $AP_{50}$ scores of 0.712, 0.864, and 0.936, and $AP_{50:95}$ scores of 0.435, 0.662, and 0.611. Notably, in terms of small object detection ($AP_{S}$), compared with other state-of-the-art models, our method achieves significant improvements of $3.5\%--14.7\%$, $2.3\%--14.2\%$, and $6.6\%--27.0\%$ on these datasets, respectively. In terms of efficiency, the model has approximately 3.5M parameters, a computational cost of about 9.0G FLOPs, and achieves real-time inference at 58.4 FPS on the Jetson Orin Nano, demonstrating its deployment potential in resource-constrained scenarios.