Gerhard P. Hancke’s research while affiliated with Nanjing University of Posts and Telecommunications and other places

Neural Radiance Fields (NeRFs) have shown remarkable performances in producing novel-view images from high-quality scene images. However, hand-held low-light photography challenges NeRFs as the captured images may simultaneously suffer from low visibility, noise, and camera shakes. While existing NeRF methods may handle either low light or motion, directly combining them or incorporating additional image-based enhancement methods does not work as these degradation factors are highly coupled. We observe that noise in low-light images is always sharp regardless of camera shakes, which implies an implicit order of these degradation factors within the image formation process. To this end, we propose in this paper a novel model, named LuSh-NeRF, which can reconstruct a clean and sharp NeRF from a group of hand-held low-light images. The key idea of LuSh-NeRF is to sequentially model noise and blur in the images via multi-view feature consistency and frequency information of NeRF, respectively. Specifically, LuSh-NeRF includes a novel Scene-Noise Decomposition (SND) module for decoupling the noise from the scene representation and a novel Camera Trajectory Prediction (CTP) module for the estimation of camera motions based on low-frequency scene information. To facilitate training and evaluations, we construct a new dataset containing both synthetic and real images. Experiments show that LuSh-NeRF outperforms existing approaches. Our code and dataset can be found here: https://github.com/quzefan/LuSh-NeRF.

Long-Range Wide-Area Network (LoRaWAN) technology offers efficient connectivity for numerous end devices over a wide coverage area in the Internet of Things (IoT) network, enabling the exchange of data over the Internet between even the most minor Internet-connected devices and systems. One of LoRaWAN’s hallmark features is the Adaptive Data Rate (ADR) algorithm. ADR is a resource allocation function which dynamically adjusts the network’s data rate, airtime, and energy dissipation to optimise its performance. The allocation of spreading factors plays a critical function in defining the throughput of the end device and its robustness to interference. However, in practical deployments, LoRaWAN networks experience considerable interference, severely affecting the packet delivery ratio, energy utilisation, and general network performance. To address this, we present a novel ADR framework, SSFIR-ADR, which utilises randomised spreading factor allocation to minimise energy consumption and packet collisions while maintaining optimal network performance. We implement a LoRa network composed of a single gateway that connects loads of end nodes to a network server. In terms of energy use, packet delivery rate, and interference rate (IR), our simulation implementation does better than LoRaWAN’s legacy ADR scheme for a range of application data intervals.

This paper explores the weakly-supervised referring image segmentation (WRIS) problem, and focuses on a challenging setup where target localization is learned directly from image-text pairs. We note that the input text description typically already contains detailed information on how to localize the target object, and we also observe that humans often follow a step-by-step comprehension process (\ie, progressively utilizing target-related attributes and relations as cues) to identify the target object. Hence, we propose a novel Progressive Comprehension Network (PCNet) to leverage target-related textual cues from the input description for progressively localizing the target object. Specifically, we first use a Large Language Model (LLM) to decompose the input text description into short phrases. These short phrases are taken as target-related cues and fed into a Conditional Referring Module (CRM) in multiple stages, to allow updating the referring text embedding and enhance the response map for target localization in a multi-stage manner. Based on the CRM, we then propose a Region-aware Shrinking (RaS) loss to constrain the visual localization to be conducted progressively in a coarse-to-fine manner across different stages. Finally, we introduce an Instance-aware Disambiguation (IaD) loss to suppress instance localization ambiguity by differentiating overlapping response maps generated by different referring texts on the same image. Extensive experiments show that our method outperforms SOTA methods on three common benchmarks.

Near-Field Communication (NFC) has enabled mobile devices to emulate contactless smart cards, which has also rendered them susceptible to relay attacks. Numerous countermeasures have been proposed that use ambient sensors as an anti-relay mechanism. However, there are concerns regarding their efficacy in time-critical scenarios, such as transport ticketing and contactless payments. This paper empirically and comprehensively evaluates whether ambient sensors are an effective anti-relay mechanism for such NFC-based contactless transactions. To this end, we examine 17 sensors available via the Android platform. Each sensor, where feasible, was used to record measurements in 1,000 contactless transactions with 252 users across four physical locations. We then conduct an extensive four-part evaluation using similarity metrics, traditional machine learning models, and deep learning methods used in existing work and beyond. We conclude that mobile ambient sensors are currently unsuitable for detecting relay attacks on NFC contactless transactions under realistic timing constraints, contrary to the suggestions and proposals made in existing work.

In 3D modeling, designers often use an existing 3D model as a reference to create new ones. This practice has inspired the development of Phidias, a novel generative model that uses diffusion for reference-augmented 3D generation. Given an image, our method leverages a retrieved or user-provided 3D reference model to guide the generation process, thereby enhancing the generation quality, generalization ability, and controllability. Our model integrates three key components: 1) meta-ControlNet that dynamically modulates the conditioning strength, 2) dynamic reference routing that mitigates misalignment between the input image and 3D reference, and 3) self-reference augmentations that enable self-supervised training with a progressive curriculum. Collectively, these designs result in a clear improvement over existing methods. Phidias establishes a unified framework for 3D generation using text, image, and 3D conditions with versatile applications.

Internet of Things (IoT) enabled WBAN systems play a crucial role in healthcare monitoring, enhancing patients’ well-being at an affordable cost. However, their primary challenge lies in energy constraints. To optimize the limited energy resources and extend the lifetime of devices within the network, a multi-channel MAC protocol with Markov decision process (MDP-HYMAC) is proposed. This protocol improves energy efficiency, throughput, and network lifetime, and minimizes delays by using separate channels for communication between biomedical devices and access points (APs). In addition, a Markov decision process is employed to stochastically model the systems’ transition states and explore optimal communication strategies between biomedical devices and APs. Furthermore, an adaptive power allocation scheme, a time-slot allocation scheme, and a back-off strategy are designed to minimize time-slot wastage, energy consumption, and delays. The proposed protocol outperforms the baseline methods by achieving significant improvements in energy efficiency from 4% to 22%. The findings of this research strongly indicate that the proposed protocol has the potential to significantly improve the performance of WBAN systems, particularly in the context of sustainable healthcare monitoring.

Privacy disclosures and malevolent data intrusions targeting adversarial agents pose significant menaces to cyber-physical systems, a reality that extends to the intricate realm of micro-grid energy management. This paper proposes a homomorphic encryption based resilient distributed algorithm with an event-triggered mechanism to address this problem. Due to the potential information disclosure issue, exchange information is encrypted to an arbitrary neighbor and decrypted with a private key to protect agents. Considering the potential security attacks on adversary agents, an event-trigger based resilient distributed optimization with trusted agents (ETRDO-T) is proposed. It ensures the convergence of distributed algorithms, as well as relives the communication burden caused by homomorphic encryption. The simulation results, it can be seen that even under data attacks from malicious nodes, this method can effectively protect privacy information in information exchange while ensuring the convergence of energy management.