This article is concerned with the data-driven health state assessment task unknown scenarios. Unknown scenarios are inaccessible in stage of model training but they can appear unexpectedly during running stage. A new problem called within-class distribution mismatch raised by assuming that still belong to known classes. To tackle this challenging problem, a novel active incremental learning scheme classifier and an out-of-distribution (OOD) detector proposed. <italic...

Real-time safety assessment of the complex dynamic systems in nonstationary environments is great significance for avoiding potential hazards. In this case, update procedure with high accuracy and training speed crucial meaningful streaming setting. Generally, performance most online learning approaches will be negatively affected by limited annotated samples such a Moreover, time cost advanced conventional methods retaining procedures relatively high, constraining their practicality....

The processing of underwater images can vastly ease the difficulty robots’ tasks and promote ocean exploration development. This paper proposes a fast efficient image enhancement model based on conditional GAN with good generalization ability using aggregation strategies concatenate operations to take full advantage limited hierarchical features. A sequential network avoid frequently visiting additional nodes, which is beneficial for speeding up inference reducing memory consumption. Through...

The underwater hyperspectral imager enables the detection and identification of targets on seafloor by collecting high-resolution spectral images. distance between cannot be consistent in real operation factors such as motion fluctuating terrain, resulting unfocused images negative effects identification. In this paper, we developed a novel integrated imaging system for deep sea surveys proposed an autofocus strategy based liquid lens focusing transfer. calibration tests provided clear focus...

Considering the inherent unknown risks in operation of complex dynamic system, it is very important to assess its safety accurately. Within framework human-in-the-loop, we propose a novel evidential group interaction-based (EGIB) assessment approach based on evidence interaction. With analysis levels (SALs), basic belief assignment (BBA) distribution multiple indicators can be obtained. Fuzzy ranking technology used collect individual's cognition these indicators. Using Shapley function...

In this work, a novel semi-supervised twin neural network (semiTNN)-based post-equalizer (PE) is proposed and developed for high-speed 4-level pulse amplitude modulation (PAM4) underwater wireless optical communication (UWOC). Furthermore, semiTNN enabled by dual-drop strategy interleaved consistency regularization (ICR), thereby only requiring small amount of labeled PAM4 symbols as references to be well-trained. The experimental results over 14-m/4-Gbps link reveal that the PE paired with...

To further leverage the technical advantages of using manned submersibles to conduct deep-sea habitat observations and their significant role in environmental monitoring protection, this paper provides an overview representative technologies for observing key elements habitats, including seabed micro-morphology, physical environment, chemical environment. These precise are essential research. The discussed encompass high-resolution acoustic optical terrain geomorphology during dives;...

Zhang, Y.; Ding, Z.; Wang, Zhao, Q., and Qin, S., 2019. Determining the damage mechanisms for buoyancy materials of deep-sea manned submersibles. Journal Coastal Research, 35(5), 996–1002. Coconut Creek (Florida), ISSN 0749-0208.Deep-sea submersibles are important platforms ocean exploration, their structural reliability is critical to personal safety crew. In particular, vital parts because they provide necessary submersible operations. This paper first describes overall used in Jiaolong...

The fault detection of manned submersibles plays a very important role in protecting the safety submersible equipment and personnel. However, diving sensor data is scarce high-dimensional, so this paper proposes method, which made up feature selection module based on hierarchical clustering Autoencoder (AE), improved Deep Convolutional Generative Adversarial Networks (DCGAN)-based augmentation using Neural Network (CNN) with LeNet-5 structure. First, developed to select features that have...

In view of the shortage structural defect monitoring methods for deep submersibles, numerical simulation and experimental research on underwater SAW propagation based interdigital transducers are carried out in this paper. PVDF transducer (PVDF-IDT) has shown considerable potential application health because its micro size, soft material characteristics, characteristics long-term bonding surface tested structure. order to realize IDT submersible or structures, it is necessary understand...

As a nonverbal cue, gaze plays critical role in communication, expressing emotions and reflecting mental activity. It has widespread applications various fields. Recently, the appearance-based estimation method, which utilizes CNN (convolutional neural networks), rapidly improved accuracy robustness of algorithms. Due to their insufficient ability capture global relationships, present methods unconstrained environments, potential for improvement. To address this challenge, focus paper is...

In this article, a spatial optical equalization method based on the Engineered Diffuser (ED) is proposed to mitigate nonlinearity degradation caused by dead time of silicon photomultiplier (SiPM). The ED can generate uniform light spot and prevent part pixels in SiPM from saturation, allowing them operate linear range. Simulation results demonstrate that reduce count loss enhance dynamic range communication. Experiments water tank are also conducted verify feasibility effectiveness...

This paper focused on the influence of initial imperfection critical load PMMA pressure spherical shell by introducing buckling modes as defect. Four types shells including complete shell, with single penetration, double penetrations, and hyperspherical were studied. The distribution imperfections was given through modal configuration linear eigenvalue analysis, then analyzed using consistent defect method N-order method. Results confirmed that: Initial have a great different varying...

Objective: Due to the problems of light propagation underwater, such as scattering and absorption, which leads low contrast, color distortion blurring details in images obtained this phenomenon is more serious deep sea, most undersea map collected by manned submersibles during deep-sea exploration exhibit these issues. To address problems, an underwater image fusion enhancement algorithm based on correction sharpening proposed for restoration. Methods: The automatic used correct enhance...

In the realm of point cloud registration, most prevalent pose evaluation approaches are statistics-based, identifying optimal transformation by maximizing number consistent correspondences. However, registration recall decreases significantly when clouds exhibit a low overlap rate, despite efforts in designing feature descriptors and establishing this paper, we introduce Deep-PE, lightweight, learning-based evaluator designed to enhance accuracy selection, especially challenging scenarios...