In recent years, network traffic analysis solutions that are driven by artificial intelligence models have achieved impressive performance. The “magic spells” of these solutions come from the knowledge that they learn from large amounts of network traffic data. However, these solutions neglect the impact of the real-world network's complexity on data quality, which makes the knowledge they learn from regular network traffic data difficult to be effective on low-quality data. Considering the packet loss in real-world network environments, this paper presents TFMana to calibrate the inaccurate packet length features extracted from incomplete network traffic data. TFMana utilizes an encoder-based masked language model to predict features of lost packets, incorporating network traffic feature embeddings to enhance prediction accuracy. This approach enables the calibrated features to approximate those extracted from loss-free network traffic asymptotically. Comprehensive experiments are conducted to verify the effectiveness of the proposed method. The evaluation demonstrates that TFMana's calibration achieves recovery accuracy between 83.94 % and 85.66 %, with minimal sensitivity to packet loss rates. Integrated with four benchmark application identification models, TFMana significantly improves classification accuracy under packet loss conditions. Notably, the analysis models maintains reliable performance even at high packet loss rates of 30 %.

Network traffic classification is critical for ensuring network quality, security, and stability. However, the increasing complexity of network environments and the growth of encrypted traffic bring significant challenges. Traditional rule-based, machine learning-based, and deep learning-based approaches are limited by the scarcity of plaintext, reliance on handcrafted features, and the need for large labeled datasets. Pre-training methods have alleviated these issues, but existing models mainly focus on payload semantics and lack dedicated learning of traffic behavior patterns essential for encrypted traffic characterization. Motivated by this, we propose TrafficBM, a dual-modality pre-training framework that jointly models semantic features and traffic behavior patterns. Our approach extracts dualmodality features from network traffic and applies modalityspecific data augmentation to mitigate data imbalance and scarcity. During pre-training, BERT leverages masked bigram modeling (MBM) to capture semantic information, while Mamba uses a masked autoencoder (MAE) architecture to learn traffic behavior patterns. An adaptive gating network, together with a parameter-preserving warm-up strategy, fuses features from both pre-trained models during fine-tuning to improve downstream classification performance. TrafficBM achieves state-of-the-art results on six tasks across eight datasets, including over 0.99 accuracy on five datasets and a 10 % improvement over the best baseline on Datacon2021 Part 2, demonstrating strong generalization and robustness in network traffic classification.

In the task of semantic understanding of remote sensing images, most current research focuses on learning contextual information through attention mechanisms or multiple inductive biases. However, these methods are limited in capturing fine-grained differences within the same attribute, are susceptible to background noise interference, and lack effective modeling capabilities for spatial relationships and long-range dependencies between different remote sensing attributes. To address these issues, we specifically focus on the homogeneous and heterogeneous differences between attributes in remote sensing images. Thus, we propose an innovative comprehensive attribute difference attention network (CADANet) to enhance the performance of understanding remote sensing images. Specifically, we design two key modules: the attribute feature aggregation (AFA) module and the context attribute-aware spatial attention (CAASA) module. The AFA module primarily focuses on global and local domain attribute modeling, reducing the impact of homogeneous attribute differences through fine-grained feature extraction and global context information. The CAASA module integrates pixel-level global background information and relative position priors, employing a self-attention mechanism to capture long-range dependencies, thus addressing heterogeneous attribute differences. Extensive experimental results conducted on the widely used Vaihingen, Potsdam, and WHDLD datasets effectively demonstrate that our proposed method outperforms other recent approaches in performance. Our code is available athttps://github.com/lzp-lkd/CADANet.

The multiattribute scene understanding (MASU) tasks currently lie in capturing multiple attribute features and learning the complex correlations between different attributes. Traditional methods primarily focus on exploring multiscale local insights and employ direct approaches to merge global semantic data into the image models, thereby neglecting the full spectrum of global semantic features across different receptive fields. Furthermore, encapsulating a wide range of spatial details through deeper networks inevitably leads to a drastic increase in computational complexity. To address these challenges, we propose a novel end-to-end network named adjacent-atrous mechanism for expanding global receptive fields (AMEGRF-Net). Specifically, we introduce an efficient local-global feature learning paradigm that innovatively expands the model’s receptive field to enhance scene understanding without incurring additional computational overhead. A local feature sensing (LFS) module is proposed to enhance the distinctiveness between different categories within the feature space while refining the spatial feature learning capability and interchannel synergy. We present an innovative adjacent-atrous mechanism, adjacent-atrous global context modeling module (AGCM), to combine a broader global receptive field with a complex relationship capturing mechanism, achieving deep modeling of the intricate relationships between attributes and labels. Through extensive comparative experiments on three challenging public datasets, the superior performance of AEGRF-Net in handling high-resolution remote sensing images for MASU has been clearly demonstrated.

In this paper, we propose a novel visible light positioning (VLP) scheme under multi-path propagation, which is a practical scenario that has not been well studied for VLP. The new scheme exploits the so-called dual-mode light-emitting diode (DM-LED) of different radiation lobe mode numbers at the transmitter and a photodiode at the receiver. Specifically, we devise a method that utilizes a radiation angle measurement approach for DM-LED, derive the Cramér-Rao lower bound (CRLB) of the estimated distance and analyze the characteristics of key parameters. In addition, two localization algorithms based on linear and nonlinear least squares methods are developed. Simulation results show that under the ideal light-of-sight scenario, the CRLB of the proposed VLP scheme can be close to that of the conventional received signal strength (RSS) aided VLP scheme. Furthermore, we implement the first ever prototype DM-LED lamp to validate the new DM-LED aided VLP system model. Both simulation and experiment results demonstrate that the proposed system outperforms its RSS-aided VLP counterpart in terms of positioning accuracy in the more realistic, and thus more challenging multi-path scenario, even with a tilting receiver.

In this paper, a novel visible light positioning (VLP) system is proposed. The new system includes multiple dual-mode light-emitting diode (DM-LED) lamps with different radiation lobe mode numbers (RLMN) and a user device equipped with a photodiode (PD). Specifically, a method utilizing a radiation angle measurement (RAM) approach for DM-LED is devised, and the characteristics of key system parameters are studied, taking into account multi-path reflections in practical indoor VLP scenarios. Furthermore, two localization algorithms based on linear and nonlinear least squares methods are developed. Simulation results show that the proposed system significantly outperforms its conventional counterpart in terms of positioning accuracy under the challenging multi-path scenario.

Indoor location based service (LBS) has become one of the key enablers for the mobile internet era. The challenging indoor localisation problem may be solved by the emerging visible light positioning (VLP) techniques which exploit light-emitting diode (LED) to transmit beacon signals. In this paper, we propose a general VLP system which invokes an imaging receiver, an off-the-shelf LED lamp and a commercial user equipment (UE) employing a rolling shutter aided complementary metal-oxide semiconductor (CMOS) based image sensor (CIS). Particularly, we design an asynchronous framework based on Miller coding, which efficiently solves the synchronisation problem. Furthermore, we provide a detailed analysis on a discrete Fourier transform (DFT) aided LED flicker frequency detection algorithm. Extensive measurement and simulation results are also offered.

Visible light communication-based schemes utilizing LED identifiers are among the most popular candidate solutions for indoor localization applications. In this paper, we design a comprehensive imaging visible light positioning system, which exploits off-the-shelf LED lamps and commercial user equipment employing a rolling shutter aided CMOS image sensor. More specifically, we first introduce an asynchronous oversampled multi-image detection scheme inspired by the rationale of Miller coding, which efficiently solves the synchronization problem in transmissions of LED identifiers. Then, a discrete Fourier transform aided LED flicker frequency detection algorithm is detailed for robust single-image detection. Furthermore, we extend the proposed method to a space-time-multiplexing framework, which improves the overall transmission rate and solves the problem of detecting the user's moving direction. The notable advantages of the new solution are demonstrated through both practical measurements and computer simulations, exhibiting a robust transmit distance beyond three meters for continuous frames.

Global soil moisture is one of the critical land surface initial conditions for numerical weather, climate, and hydrological predictions. Since it is not practical to provide global maps using ground measurements, land surface soil moisture remote sensing has been a hot research topic in the last several decades. As a result, a number of soil moisture products have been produced from different satellite sensors with different spatial and temporal coverage and quality. To make effective use of all available soil moisture products, a Soil Moisture Operational Product System (SMOPS) has been developed at National Oceanic and Atmospheric Administration (NOAA) to produce a one-stop shop for all operational soil moisture products from different satellite sensors. To increase the spatial and temporal coverage of soil moisture product, SMOPS also provides a data layer that merges soil moisture retrievals from multiple satellites in addition to the individual soil moisture retrievals from each of the available satellites. This paper gives an overall introduction of SMOPS and its validations using ground measurements.