[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
7.3
3.4
Journals
Sensors
Special Issues
Advances in Deep Learning for Intelligent Sensing Systems
sensors-logo
Submit to Sensors Review for Sensors Propose a Special Issue

Journal Menu

► Journal Menu

Journal Browser

► Journal Browser
share announcement

Advances in Deep Learning for Intelligent Sensing Systems

A special issue of Sensors (ISSN 1424-8220). This special issue belongs to the section "Intelligent Sensors".

Deadline for manuscript submissions: closed (30 November 2022) | Viewed by 38368

Special Issue Editors

Dr. Min Xia

E-Mail Website
Guest Editor
Department of Engineering, Lancaster University, Lancaster LA1 4YW, UK
Interests: machine condition monitoring; smartning machine condition monitoring; smart sensing; machine learning
Special Issues, Collections and Topics in MDPI journals
Prof. Dr. Teng Li

E-Mail Website
Guest Editor
School of Control Science and Engineering, Shandong University, Jinan, China
Interests: machine learning; intelligent systems; wind farms, robotics; their applications in structural and machine health monitoring
Special Issues, Collections and Topics in MDPI journals
Prof. Dr. Clarence de Silva

E-Mail Website
Guest Editor
Department of Mechanical Engineering, University of British Columbia, Vancouver, BC V6T 1Z4, Canada
Interests: mechatronics; sensors and actuators; intelligent control; robotics

Special Issue Information

Dear Colleagues,

Recent advances in deep learning techniques have led to significant progress in sensing systems. Today, intelligent sensing can benefit sensing processes in many research and application fields. For example, inherent and complex time-series or spatiotemporal correlations among sensing data can be exploited using deep learning to characterize the studied objectives of interest. In many real-world sensing scenarios, the high-level features within sensing data imply underlying or unknown interactions leading to useful information or knowledge, which may outperform traditional analytical tools when learning and interpreting studied problems.

The Special Issue focuses on advanced deep learning methods, e.g., deep neural networks, to address a broad view of problems about sensor, sensing, and sensory issues in intelligent sensing systems. Contributions are encouraged to design and develop novel deep learning frameworks, particularly concentrating on their effectiveness, intelligence, and reliability in solving challenging sensing issues or achieving superior sensing performance. This Special Issue is dedicated to both theoretical innovations and real-world applications with field implementation and experiments. The topics of this issue include but are not limited to the following:

  • Data acquisition, analysis, and decision making in sensors, robotics, and intelligent systems;
  • Deep neural networks, deep attention mechanisms, and other representative learning techniques;
  • Classification, regression, interpretation, and prediction for intelligent sensing;
  • Machine condition monitoring, environmental monitoring, structural health monitoring, and other monitoring programs;
  • Comparison studies between traditional methods and advanced deep learning methods in sensing problems.

Dr. Min Xia
Prof. Dr. Teng Li
Prof. Dr. Clarence de Silva
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Sensors is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2600 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

  • Deep learning
  • Deep neural network
  • Sensor and sensing system
  • Intelligent sensing
  • Big data analysis
  • Decision making

Benefits of Publishing in a Special Issue

  • Ease of navigation: Grouping papers by topic helps scholars navigate broad scope journals more efficiently.
  • Greater discoverability: Special Issues support the reach and impact of scientific research. Articles in Special Issues are more discoverable and cited more frequently.
  • Expansion of research network: Special Issues facilitate connections among authors, fostering scientific collaborations.
  • External promotion: Articles in Special Issues are often promoted through the journal's social media, increasing their visibility.
  • e-Book format: Special Issues with more than 10 articles can be published as dedicated e-books, ensuring wide and rapid dissemination.

Further information on MDPI's Special Issue policies can be found here.

Published Papers (9 papers)

Download All Papers
Order results
Result details
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:

Research

Jump to: Review

13 pages, 1602 KiB  
Article
Text Summarization Method Based on Gated Attention Graph Neural Network
by Jingui Huang, Wenya Wu, Jingyi Li and Shengchun Wang
Sensors 2023, 23(3), 1654; https://doi.org/10.3390/s23031654 - 2 Feb 2023
Cited by 6 | Viewed by 2753
Abstract
Text summarization is an information compression technology to extract important information from long text, which has become a challenging research direction in the field of natural language processing. At present, the text summary model based on deep learning has shown good results, but [...] Read more.
Text summarization is an information compression technology to extract important information from long text, which has become a challenging research direction in the field of natural language processing. At present, the text summary model based on deep learning has shown good results, but how to more effectively model the relationship between words, more accurately extract feature information and eliminate redundant information is still a problem of concern. This paper proposes a graph neural network model GA-GNN based on gated attention, which effectively improves the accuracy and readability of text summarization. First, the words are encoded using a concatenated sentence encoder to generate a deeper vector containing local and global semantic information. Secondly, the ability to extract key information features is improved by using gated attention units to eliminate local irrelevant information. Finally, the loss function is optimized from the three aspects of contrastive learning, confidence calculation of important sentences, and graph feature extraction to improve the robustness of the model. Experimental validation was conducted on a CNN/Daily Mail dataset and MR dataset, and the results showed that the model in this paper outperformed existing methods. Full article
(This article belongs to the Special Issue Advances in Deep Learning for Intelligent Sensing Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Model framework based on gated graph attention network.</p> Full article ">Figure 2
<p>Gated Attention Unit GAU.</p> Full article ">Figure 3
<p>Experimental results of the effect of different modules on model performance.</p> Full article ">
22 pages, 931 KiB  
Article
Novel Meta-Learning Techniques for the Multiclass Image Classification Problem
by Antonios Vogiatzis, Stavros Orfanoudakis, Georgios Chalkiadakis, Konstantia Moirogiorgou and Michalis Zervakis
Sensors 2023, 23(1), 9; https://doi.org/10.3390/s23010009 - 20 Dec 2022
Cited by 1 | Viewed by 3407
Abstract
Multiclass image classification is a complex task that has been thoroughly investigated in the past. Decomposition-based strategies are commonly employed to address it. Typically, these methods divide the original problem into smaller, potentially simpler problems, allowing the application of numerous well-established learning algorithms [...] Read more.
Multiclass image classification is a complex task that has been thoroughly investigated in the past. Decomposition-based strategies are commonly employed to address it. Typically, these methods divide the original problem into smaller, potentially simpler problems, allowing the application of numerous well-established learning algorithms that may not apply directly to the original task. This work focuses on the efficiency of decomposition-based methods and proposes several improvements to the meta-learning level. In this paper, four methods for optimizing the ensemble phase of multiclass classification are introduced. The first demonstrates that employing a mixture of experts scheme can drastically reduce the number of operations in the training phase by eliminating redundant learning processes in decomposition-based techniques for multiclass problems. The second technique for combining learner-based outcomes relies on Bayes’ theorem. Combining the Bayes rule with arbitrary decompositions reduces training complexity relative to the number of classifiers even further. Two additional methods are also proposed for increasing the final classification accuracy by decomposing the initial task into smaller ones and ensembling the output of the base learners along with that of a multiclass classifier. Finally, the proposed novel meta-learning techniques are evaluated on four distinct datasets of varying classification difficulty. In every case, the proposed methods present a substantial accuracy improvement over existing traditional image classification techniques. Full article
(This article belongs to the Special Issue Advances in Deep Learning for Intelligent Sensing Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Ensemble learning voting architecture.</p> Full article ">Figure 2
<p>Stacked generalization architecture.</p> Full article ">Figure 3
<p>The architecture of the Bayes-theorem-based Ensemble approach.</p> Full article ">Figure 4
<p>Mixture of experts; one-vs.-rest classification.</p> Full article ">Figure 5
<p>Meta-Learning, binary classifiers only.</p> Full article ">Figure 6
<p>Meta-Learning, binary, and multiclass classifiers generalizer.</p> Full article ">Figure 7
<p>Network architecture of the <span class="html-italic">generalizers</span>.</p> Full article ">
20 pages, 1384 KiB  
Article
Defect Severity Identification for a Catenary System Based on Deep Semantic Learning
by Jian Wang, Shibin Gao, Long Yu, Dongkai Zhang and Lei Kou
Sensors 2022, 22(24), 9922; https://doi.org/10.3390/s22249922 - 16 Dec 2022
Cited by 1 | Viewed by 2062
Abstract
A variety of Chinese textual operational text data has been recorded during the operation and maintenance of the high-speed railway catenary system. Such defect text records can facilitate defect detection and defect severity analysis if mined efficiently and accurately. Therefore, in this context, [...] Read more.
A variety of Chinese textual operational text data has been recorded during the operation and maintenance of the high-speed railway catenary system. Such defect text records can facilitate defect detection and defect severity analysis if mined efficiently and accurately. Therefore, in this context, this paper focuses on a specific problem in defect text mining, which is to efficiently extract defect-relevant information from catenary defect text records and automatically identify catenary defect severity. The specific task is transformed into a machine learning problem for defect text classification. First, we summarize the characteristics of catenary defect texts and construct a text dataset. Second, we use BERT to learn defect texts and generate word embedding vectors with contextual features, fed into the classification model. Third, we developed a deep text categorization network (DTCN) to distinguish the catenary defect level, considering the contextualized semantic features. Finally, the effectiveness of our proposed method (BERT-DTCN) is validated using a catenary defect textual dataset collected from 2016 to 2018 in the China Railway Administration in Chengdu, Lanzhou, and Hengshui. Moreover, BERT-DTCN outperforms several competitive methods in terms of accuracy, precision, recall, and F1-score value. Full article
(This article belongs to the Special Issue Advances in Deep Learning for Intelligent Sensing Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Scene of a catenary system.</p> Full article ">Figure 2
<p>Development process of text classification.</p> Full article ">Figure 3
<p>Architecture of defect severity identification for a catenary system.</p> Full article ">Figure 4
<p>The content of the catenary defect record.</p> Full article ">Figure 5
<p>Structure of the BERT model.</p> Full article ">Figure 6
<p>An example of the BERT input representation.</p> Full article ">Figure 7
<p>Basic structure of the transformer encoder.</p> Full article ">Figure 8
<p>The flowchart of the self-attention layer.</p> Full article ">Figure 9
<p>The training loss of DTCN and BERT-DTCN.</p> Full article ">Figure 10
<p>The ROC curves for DTCN and BERT-DTCN.</p> Full article ">Figure 11
<p>The PR curves for DTCN and BERT-DTCN.</p> Full article ">Figure 12
<p>The training loss of BERT-DTCN and competing models.</p> Full article ">Figure 13
<p>The ROC curves for BERT-DTCN and competing models.</p> Full article ">Figure 14
<p>The PR curves for BERT-DTCN and competing models.</p> Full article ">
13 pages, 1450 KiB  
Article
Dimension Reduction of Digital Image Descriptors in Neural Identification of Damaged Malting Barley Grains
by Piotr Boniecki, Agnieszka Sujak, Agnieszka A. Pilarska, Hanna Piekarska-Boniecka, Agnieszka Wawrzyniak and Barbara Raba
Sensors 2022, 22(17), 6578; https://doi.org/10.3390/s22176578 - 31 Aug 2022
Cited by 5 | Viewed by 1717
Abstract
The paper covers the problem of determination of defects and contamination in malting barley grains. The analysis of the problem indicated that although several attempts have been made, there are still no effective methods of identification of the quality of barley grains, such [...] Read more.
The paper covers the problem of determination of defects and contamination in malting barley grains. The analysis of the problem indicated that although several attempts have been made, there are still no effective methods of identification of the quality of barley grains, such as the use of information technology, including intelligent sensors (currently, quality assessment of grain is performed manually). The aim of the study was the construction of a reduced set of the most important graphic descriptors from machine-collected digital images, important in the process of neural evaluation of the quality of BOJOS variety malting barley. Grains were sorted into three size fractions and seed images were collected. As a large number of graphic descriptors implied difficulties in the development and operation of neural classifiers, a PCA (Principal Component Analysis) statistical method of reducing empirical data contained in the analyzed set was applied. The grain quality expressed by an optimal set of transformed descriptors was modelled using artificial neural networks (ANN). The input layer consisted of eight neurons with a linear Postsynaptic Function (PSP) and a linear activation function. The one hidden layer was composed of sigmoid neurons having a linear PSP function and a logistic activation function. One sigmoid neuron was the output of the network. The results obtained show that neural identification of digital images with application of Principal Component Analysis (PCA) combined with neural classification is an effective tool supporting the process of rapid and reliable quality assessment of BOJOS malting barley grains. Full article
(This article belongs to the Special Issue Advances in Deep Learning for Intelligent Sensing Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>The scheme of the proposed procedures.</p> Full article ">Figure 2
<p>Examples of the types of BOJOS cultivar grain damages.</p> Full article ">Figure 3
<p>Acquisition and image processing of damaged malting barley grains using the original Hordeum v. 3.2 computer system created by B. Raba within MATLAB 2014b environment (MathWorks, Natick, MA, USA) using Image Processing Toolbox library.</p> Full article ">
18 pages, 5598 KiB  
Article
Few-Shot Text Classification with Global–Local Feature Information
by Depei Wang, Zhuowei Wang, Lianglun Cheng and Weiwen Zhang
Sensors 2022, 22(12), 4420; https://doi.org/10.3390/s22124420 - 11 Jun 2022
Cited by 3 | Viewed by 2152
Abstract
Meta-learning frameworks have been proposed to generalize machine learning models for domain adaptation without sufficient label data in computer vision. However, text classification with meta-learning is less investigated. In this paper, we propose SumFS to find global top-ranked sentences by extractive summary and [...] Read more.
Meta-learning frameworks have been proposed to generalize machine learning models for domain adaptation without sufficient label data in computer vision. However, text classification with meta-learning is less investigated. In this paper, we propose SumFS to find global top-ranked sentences by extractive summary and improve the local vocabulary category features. The SumFS consists of three modules: (1) an unsupervised text summarizer that removes redundant information; (2) a weighting generator that associates feature words with attention scores to weight the lexical representations of words; (3) a regular meta-learning framework that trains with limited labeled data using a ridge regression classifier. In addition, a marine news dataset was established with limited label data. The performance of the algorithm was tested on THUCnews, Fudan, and marine news datasets. Experiments show that the SumFS can maintain or even improve accuracy while reducing input features. Moreover, the training time of each epoch is reduced by more than 50%. Full article
(This article belongs to the Special Issue Advances in Deep Learning for Intelligent Sensing Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>The SumFS framework. The blue arrow denotes the global feature processing process, the black arrow denotes the training process, and the red arrow denotes the testing process.</p> Full article ">Figure 2
<p>CH index under several Sum-H. The number of sentences corresponding to the red points was chosen.</p> Full article ">Figure 3
<p>TSNE visualization of the input representations for Sum-H and raw data of three datasets, which are THUCnews, Fudan news and Marine news. Their Sum-H’s distributions are shown in (<b>a</b>), (<b>c</b>) and (<b>e</b>), respectively, and their distributions of raw data are shown in (<b>b</b>), (<b>d</b>) and (<b>f</b>), respectively. Each color/marker pair corresponds to a specific label. The class numbers are listed in <a href="#sensors-22-04420-t001" class="html-table">Table 1</a>.</p> Full article ">Figure 3 Cont.
<p>TSNE visualization of the input representations for Sum-H and raw data of three datasets, which are THUCnews, Fudan news and Marine news. Their Sum-H’s distributions are shown in (<b>a</b>), (<b>c</b>) and (<b>e</b>), respectively, and their distributions of raw data are shown in (<b>b</b>), (<b>d</b>) and (<b>f</b>), respectively. Each color/marker pair corresponds to a specific label. The class numbers are listed in <a href="#sensors-22-04420-t001" class="html-table">Table 1</a>.</p> Full article ">Figure 4
<p>Comparison of IDF-IWF-ATT results with different numbers of sentences.</p> Full article ">Figure 5
<p>The accuracy trends of training and testing using the ATT-IDF-IWF weighted strategy. According to the results, Sum-H preserves the original text information well.</p> Full article ">Figure 6
<p>Comparison of the standard deviations during the testing steps.</p> Full article ">Figure 7
<p>Comparison of results with the optimal solution strategy.</p> Full article ">Figure 8
<p>Comparison of time consumption between Raw and Sum-H. These results were estimated using the average time consumption for each epoch.</p> Full article ">
18 pages, 4301 KiB  
Article
Deep Reinforcement Learning Microgrid Optimization Strategy Considering Priority Flexible Demand Side
by Jinsong Sang, Hongbin Sun and Lei Kou
Sensors 2022, 22(6), 2256; https://doi.org/10.3390/s22062256 - 14 Mar 2022
Cited by 25 | Viewed by 4280
Abstract
As an efficient way to integrate multiple distributed energy resources (DERs) and the user side, a microgrid is mainly faced with the problems of small-scale volatility, uncertainty, intermittency and demand-side uncertainty of DERs. The traditional microgrid has a single form and cannot meet [...] Read more.
As an efficient way to integrate multiple distributed energy resources (DERs) and the user side, a microgrid is mainly faced with the problems of small-scale volatility, uncertainty, intermittency and demand-side uncertainty of DERs. The traditional microgrid has a single form and cannot meet the flexible energy dispatch between the complex demand side and the microgrid. In response to this problem, the overall environment of wind power, thermostatically controlled loads (TCLs), energy storage systems (ESSs), price-responsive loads and the main grid is proposed. Secondly, the centralized control of the microgrid operation is convenient for the control of the reactive power and voltage of the distributed power supply and the adjustment of the grid frequency. However, there is a problem in that the flexible loads aggregate and generate peaks during the electricity price valley. The existing research takes into account the power constraints of the microgrid and fails to ensure a sufficient supply of electric energy for a single flexible load. This paper considers the response priority of each unit component of TCLs and ESSs on the basis of the overall environment operation of the microgrid so as to ensure the power supply of the flexible load of the microgrid and save the power input cost to the greatest extent. Finally, the simulation optimization of the environment can be expressed as a Markov decision process (MDP) process. It combines two stages of offline and online operations in the training process. The addition of multiple threads with the lack of historical data learning leads to low learning efficiency. The asynchronous advantage actor–critic (Memory A3C, M-A3C) with the experience replay pool memory library is added to solve the data correlation and nonstatic distribution problems during training. The multithreaded working feature of M-A3C can efficiently learn the resource priority allocation on the demand side of the microgrid and improve the flexible scheduling of the demand side of the microgrid, which greatly reduces the input cost. Comparison of the researched cost optimization results with the results obtained with the proximal policy optimization (PPO) algorithm reveals that the proposed algorithm has better performance in terms of convergence and optimization economics. Full article
(This article belongs to the Special Issue Advances in Deep Learning for Intelligent Sensing Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Microgrid architecture.</p> Full article ">Figure 2
<p>M-A3C microgrid management structure.</p> Full article ">Figure 3
<p>Microgrid wind power generation curve.</p> Full article ">Figure 4
<p>Initial state of microgrid environment components.</p> Full article ">Figure 5
<p>Cumulative reward curves for M-A3C and PPO training.</p> Full article ">Figure 6
<p>Microgrid and main grid energy interaction.</p> Full article ">Figure 7
<p>Microgrid and TCL energy dispatch.</p> Full article ">Figure 8
<p>ESS daily charge.</p> Full article ">Figure 9
<p>Daily distribution and consumption of TCL.</p> Full article ">Figure 10
<p>A3C and PPO optimization performance comparison.</p> Full article ">Figure 11
<p>Energy interaction between PPO microgrid and main grid.</p> Full article ">
21 pages, 32597 KiB  
Article
Using Voice Activity Detection and Deep Neural Networks with Hybrid Speech Feature Extraction for Deceptive Speech Detection
by Serban Mihalache and Dragos Burileanu
Sensors 2022, 22(3), 1228; https://doi.org/10.3390/s22031228 - 6 Feb 2022
Cited by 15 | Viewed by 6622
Abstract
In this work, we first propose a deep neural network (DNN) system for the automatic detection of speech in audio signals, otherwise known as voice activity detection (VAD). Several DNN types were investigated, including multilayer perceptrons (MLPs), recurrent neural networks (RNNs), and convolutional [...] Read more.
In this work, we first propose a deep neural network (DNN) system for the automatic detection of speech in audio signals, otherwise known as voice activity detection (VAD). Several DNN types were investigated, including multilayer perceptrons (MLPs), recurrent neural networks (RNNs), and convolutional neural networks (CNNs), with the best performance being obtained for the latter. Additional postprocessing techniques, i.e., hysteretic thresholding, minimum duration filtering, and bilateral extension, were employed in order to boost performance. The systems were trained and tested using several data subsets of the CENSREC-1-C database, with different simulated ambient noise conditions, and additional testing was performed on a different CENSREC-1-C data subset containing actual ambient noise, as well as on a subset of the TIMIT database. An accuracy of up to 99.13% was obtained for the CENSREC-1-C datasets, and 97.60% for the TIMIT dataset. We proceed to show how the final VAD system can be adapted and employed within an utterance-level deceptive speech detection (DSD) processing pipeline. The best DSD performance is achieved by a novel hybrid CNN-MLP network leveraging a fusion of algorithmically and automatically extracted speech features, and reaches an unweighted accuracy (UA) of 63.7% on the RLDD database, and 62.4% on the RODeCAR database. Full article
(This article belongs to the Special Issue Advances in Deep Learning for Intelligent Sensing Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>CNN-based VAD system. In this configuration example, the three convolutional (Conv) layers use 16, 32, and 64 filters, respectively, with corresponding kernel sizes of 16, 8, and 8. The size of the max-pooling layers used after the Conv layers is 2. The third Conv layer’s output is flattened into a one-dimensional vector and fed through a fully connected hidden layer with 8 nodes, used together with the output neuron with the “sigmoid” activation function for final frame-level classification. (<b>a</b>) Block diagram representation; (<b>b</b>) Layer diagram representation using Net2Vis [<a href="#B33-sensors-22-01228" class="html-bibr">33</a>].</p> Full article ">Figure 2
<p>Hysteretic thresholding. For a window to be considered <span class="html-italic">positive</span> after the previous ones had been <span class="html-italic">negative</span>, the probability must be greater than <span class="html-italic">TH<sub>med</sub></span> + ∆<span class="html-italic">H</span>. Similarly, for the opposite case, the threshold is <span class="html-italic">TH<sub>med</sub></span> − ∆<span class="html-italic">H</span>.</p> Full article ">Figure 3
<p>(<b>a</b>) Minimum duration filtering; if a resulting utterance has a shorter duration than a reference value, ∆<span class="html-italic">t<sub>min</sub></span>, it is discarded; (<b>b</b>) Bilateral extension; the utterance start time is lowered by a value ∆<span class="html-italic">t<sub>ext</sub></span>, while its stop time is increased by the same value.</p> Full article ">Figure 4
<p>Layer diagram representation of the final hybrid CNN-MLP-based VAD system, using the same configuration example as in <a href="#sensors-22-01228-f001" class="html-fig">Figure 1</a>.</p> Full article ">Figure 5
<p>Layer diagram representation of the final hybrid CNN-MLP-based VAD system, using the same configuration example as in <a href="#sensors-22-01228-f001" class="html-fig">Figure 1</a>.</p> Full article ">Figure 6
<p>Layer diagram representation of the basic DSD MLP-based system, using between 2 and 3 hidden layers with 64 or 128 nodes per layer, and with an output layer of size 2, the number of classes taken into account (<span class="html-italic">truthful</span> and <span class="html-italic">deceptive</span>), and applying the “softmax” activation function.</p> Full article ">Figure 7
<p>Detailed block diagram representation of the preprocessing and feature extraction stages of the DSD system. The VAD system is used to detect and split the input audio into utterances from each of which the set of 2260 features is extracted.</p> Full article ">Figure 8
<p>Layer diagram representation of the CNN-based DSD system. In this configuration example, the three convolutional (Conv) layers use 16, 32, and 64 filters, respectively, with kernel sizes of 3 × 3. The size of the max-pooling layers used after the Conv layers is 2 × 2. The third Conv layer’s output is flattened into a one-dimensional vector and fed through two fully connected hidden layers with 32 nodes each, used together with the output layer with the “softmax” activation function for utterance-level classification.</p> Full article ">Figure 9
<p>CENSREC-1-C subset B2; CNN-based VAD. “High SNR” groups the 10–20 dB levels and “Low SNR” groups the −5–5 dB levels: (<b>a</b>) FRR vs. ∆<span class="html-italic">t<sub>ext</sub></span>; (<b>b</b>) accuracy vs. ∆<span class="html-italic">t<sub>ext</sub></span>.</p> Full article ">Figure 10
<p>CENSREC-1-C subset B2; CNN-based VAD. “High SNR” groups the 10–20 dB levels and “Low SNR” groups the −5–5 dB levels: accuracy vs. ∆<span class="html-italic">t<sub>min</sub></span>.</p> Full article ">Figure 11
<p>RLDD and RODeCAR; final CNN-MLP-based VAD, with expanded frequency-domain input (255-pt. DFT at 16 kHz sampling rate), retrained on 10% of the data: accuracy vs. ∆<span class="html-italic">t<sub>ext</sub></span>.</p> Full article ">
13 pages, 2372 KiB  
Article
Real-Time Closed-Loop Detection Method of vSLAM Based on a Dynamic Siamese Network
by Quande Yuan, Zhenming Zhang, Yuzhen Pi, Lei Kou and Fangfang Zhang
Sensors 2021, 21(22), 7612; https://doi.org/10.3390/s21227612 - 16 Nov 2021
Cited by 5 | Viewed by 2581
Abstract
As visual simultaneous localization and mapping (vSLAM) is easy disturbed by the changes of camera viewpoint and scene appearance when building a globally consistent map, the robustness and real-time performance of key frame image selections cannot meet the requirements. To solve this problem, [...] Read more.
As visual simultaneous localization and mapping (vSLAM) is easy disturbed by the changes of camera viewpoint and scene appearance when building a globally consistent map, the robustness and real-time performance of key frame image selections cannot meet the requirements. To solve this problem, a real-time closed-loop detection method based on a dynamic Siamese networks is proposed in this paper. First, a dynamic Siamese network-based fast conversion learning model is constructed to handle the impact of external changes on key frame judgments, and an elementwise convergence strategy is adopted to ensure the accurate positioning of key frames in the closed-loop judgment process. Second, a joint training strategy is designed to ensure the model parameters can be learned offline in parallel from tagged video sequences, which can effectively improve the speed of closed-loop detection. Finally, the proposed method is applied experimentally to three typical closed-loop detection scenario datasets and the experimental results demonstrate the effectiveness and robustness of the proposed method under the interference of complex scenes. Full article
(This article belongs to the Special Issue Advances in Deep Learning for Intelligent Sensing Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Execution flow of the real-time closed-loop detection method.</p> Full article ">Figure 2
<p>Conversion of the target appearance change.</p> Full article ">Figure 3
<p>Conversion process of the background suppression.</p> Full article ">Figure 4
<p>Architecture of the dynamic Siamese network.</p> Full article ">Figure 5
<p>Comparison of the robustness of different closed-loop detection algorithms: (<b>a</b>) Gardens Point; (<b>b</b>) Mapillary; (<b>c</b>) Nordland.</p> Full article ">Figure 6
<p>Average accuracy of the different algorithms with the different datasets.</p> Full article ">

Review

Jump to: Research

36 pages, 5109 KiB  
Review
Review on Monitoring, Operation and Maintenance of Smart Offshore Wind Farms
by Lei Kou, Yang Li, Fangfang Zhang, Xiaodong Gong, Yinghong Hu, Quande Yuan and Wende Ke
Sensors 2022, 22(8), 2822; https://doi.org/10.3390/s22082822 - 7 Apr 2022
Cited by 71 | Viewed by 10509
Abstract
In recent years, with the development of wind energy, the number and scale of wind farms have been developing rapidly. Since offshore wind farms have the advantages of stable wind speed, being clean, renewable, non-polluting, and the non-occupation of cultivated land, they have [...] Read more.
In recent years, with the development of wind energy, the number and scale of wind farms have been developing rapidly. Since offshore wind farms have the advantages of stable wind speed, being clean, renewable, non-polluting, and the non-occupation of cultivated land, they have gradually become a new trend in the wind power industry all over the world. The operation and maintenance of offshore wind power has been developing in the direction of digitization and intelligence. It is of great significance to carry out research on the monitoring, operation, and maintenance of offshore wind farms, which will be of benefit for the reduction of the operation and maintenance costs, the improvement of the power generation efficiency, improvement of the stability of offshore wind farm systems, and the building of smart offshore wind farms. This paper will mainly summarize the monitoring, operation, and maintenance of offshore wind farms, with particular focus on the following points: monitoring of “offshore wind power engineering and biological and environment”, the monitoring of power equipment, and the operation and maintenance of smart offshore wind farms. Finally, the future research challenges in relation to the monitoring, operation, and maintenance of smart offshore wind farms are proposed, and the future research directions in this field are explored, especially in marine environment monitoring, weather and climate prediction, intelligent monitoring of power equipment, and digital platforms. Full article
(This article belongs to the Special Issue Advances in Deep Learning for Intelligent Sensing Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Monitoring, operation, and maintenance system of smart offshore wind farms.</p> Full article ">Figure 2
<p>Some meteorological monitoring equipment: (<b>a</b>) BCF handheld anemometer; (<b>b</b>) Scanning aerosol lidar; (<b>c</b>) Ship meteorological instrument; (<b>d</b>) SXZ2-2 Hydrometeorological automatic observation system.</p> Full article ">Figure 3
<p>Long-term observation system of air-sea coupling in Greenland.</p> Full article ">Figure 4
<p>SBF series coastal telemetering wave gauge.</p> Full article ">Figure 5
<p>Natural disasters and environmental pollution: (<b>a</b>) sea ice; (<b>b</b>) oil spill; (<b>c</b>) storm surge; (<b>d</b>) red tide.</p> Full article ">Figure 6
<p>IOISAS Seatrix.</p> Full article ">Figure 7
<p>Underwater junction box observation network system.</p> Full article ">Figure 8
<p>Some robots with different functions: (<b>a</b>) Spraying robot; (<b>b</b>) Small diameter pipe robot; (<b>c</b>) ROV II; (<b>d</b>) ROS robot.</p> Full article ">Figure 9
<p>Downtime distribution of each part.</p> Full article ">Figure 10
<p>Gearbox fault diagnosis flowchart.</p> Full article ">Figure 11
<p>Stacking fusion algorithm framework (RF: Random Forest; SVM: Support Vector Machines; KNN: K Near Neighbor; GBDT: Gradient Boosting Decision Tree).</p> Full article ">Figure 12
<p>Short-circuit fault isolation technology with fast fuses: (<b>a</b>) Two-level; (<b>b</b>) NPC.</p> Full article ">Figure 13
<p>AI(Artificial Intelligence)-based open-circuit fault diagnosis methods(ANN: Artificial Neural Network; CNN: Convolutional Neural Networks).</p> Full article ">Figure 14
<p>Fault diagnosis schematic for power electronic energy conversion systems.</p> Full article ">Figure 15
<p>Offshore booster station.</p> Full article ">Figure 16
<p>UPS (Uninterruptible Power Supply) monitoring system.</p> Full article ">Figure 17
<p>Operation and maintenance cost of offshore wind power.</p> Full article ">Figure 18
<p>Intelligent dispatching management system of offshore wind farms.</p> Full article ">Figure 19
<p>Transportation for the operation and maintenance of smart offshore wind farms.</p> Full article ">Figure 20
<p>Main influential factors in the maintenance of offshore wind farms.</p> Full article ">Figure 21
<p>Anti-typhoon strategy for offshore wind farms.</p> Full article ">Figure 22
<p>Operation and maintenance strategy.</p> Full article ">Figure 23
<p>Search and rescue in a maritime emergency.</p> Full article ">Figure 24
<p>Operation and maintenance based on the pressure of operation and maintenance personnel.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-9
Back to TopTop