The utilization of low Earth orbit (LEO) satellites’ signals of opportunity (SOPs) for absolute positioning and navigation in global navigation satellite system (GNSS)-denied environments has emerged as a significant area of research. Among various methodologies, tightly integrated Doppler/inertial navigation system (INS) frameworks present a promising solution for achieving real-time LEO-SOP-based positioning in dynamic scenarios. However, existing integration schemes generally overlook the key characteristics of LEO opportunity signals, including the limited number of visible satellites and the random nature of signal broadcasts. These factors exacerbate the weak observability inherent in LEO-SoOP Doppler/INS positioning, resulting in difficulty in obtaining reliable solutions and degraded positioning accuracy. To address these issues, this paper proposes a novel LEO-SOP Doppler/INS tight integration method that incorporates trending information to alleviate the problem of weak observability. The method leverages a parallel filtering structure combining extended Kalman filter (EKF) and Rauch–Tung–Striebel (RTS) smoothing, extracting trend information from the quasi-real-time high-precision RTS filtering results to optimize the EKF positioning solution for the current epoch. This approach effectively avoids the overfitting problem commonly associated with directly using batch data to estimate the current epoch state. The experimental results validate the improved positioning accuracy and robustness of the proposed method.
Few-shot fine-grained (FSFG) categorization aims to classify new, fine-grained concepts using limited annotated data for each class. It requires a model capable of learning subtle, distinguishable features from similar classes. An intuitive approach is to identify several discriminative parts based on the input image and represent it using these localized parts. It is interpretable since the parts obtained and combined can be seen as the way humans learn about new concepts. Therefore, we propose a straightforward and efficient module for the FSFG task by emphasizing the importance of information from discriminative parts. Specifically, we design a plug-and-play meta-detector based on object detection models, which identifies the positions of distinguishable parts in fine-grained images. With the positions of objects and semantic parts identified, we can extract richer features to enhance few-shot results. Extensive experiments on five fine-grained datasets demonstrate that our approach is applicable to various few-shot backbones and significantly improves accuracy without adding extra parameters. More importantly, compared to other variants on the same base few-shot network, the meta-detector module significantly outperforms them, further demonstrating its effectiveness.
29 December 2025 in Circuits, Systems, and Signal Processing
Global Navigation Satellite System Reflectometry (GNSS-R) provides a promising approach for wind speed retrieval over inland waters, with relevance to wind energy assessment and lake–atmosphere exchange studies. Existing GNSS-R wind retrieval methods are well established for open oceans but face major challenges over inland waters, where coherent scattering dominates and traditional ocean models produce large systematic biases. Unlike open oceans, inland waters are dominated by coherent scattering due to limited fetch, resulting in Delay-Doppler Maps (DDM) with highly concentrated energy and minimal spreading. These characteristics render conventional ocean-based retrieval models—built on incoherent scattering assumptions—often inadequate. To overcome this, we develop a lightweight convolutional neural network (CNN) tailored to the coherent regime, using raw CYGNSS DDM as input for end-to-end wind speed regression. Cross-seasonal validation over Lake Victoria and Lake Hongze shows that the model robustly captures wind-driven spatiotemporal patterns aligned with ERA5. Notably, ERA5 reanalysis winds exhibit uncertainties over inland waters, with a root mean square error (RMSE) of 1.5–2.5 m/s against in situ buoys. The model yields a low RMSE (<0.7 m/s) in reconstructing ERA5-resolved wind patterns. This work extends GNSS-R to inland waters, offering a lightweight, deployable remote sensing solution for wind energy and lake–atmosphere research.