检索结果-内蒙古大学图书馆

Robust watermarking using diffusion of logo into auto-encoder feature maps

MULTIMEDIA TOOLS AND APPLICATIONS 2023年第29期82卷 45175-45201页

作者： Jamali, Maedeh Karimi, Nader Khadivi, Pejman Shirani, Shahram Samavi, Shadrokh Isfahan Univ Technol Dept Elect & Comp Engn Esfahan *** Iran Seattle Univ Comp Sci Dept Seattle WA 98122 USA McMaster Univ Elect & Comp Engn Dept Hamilton ON Canada

Digital content has grown dramatically in recent years, leading to increased attention to copyright. Image watermarking has been considered one of the most popular methods for copyright protection. With the recent advancements in the application of deep neural networks in image processing, these networks have also been used in image watermarking. Robustness and imperceptibility are two challenging features of watermarking methods that the trade-off between them should be satisfied. In this paper, we propose to use an end-to-end network for watermarking. We use a convolutional neural network (CNN) to control the embedding strength based on the images' content. Dynamic embedding helps the network to have the lowest effect on the visual quality of the watermarked image. Different image processing attacks are simulated as a network layer to improve the robustness of the model. Our method is a blind watermarking approach that replicates the watermark string to create a matrix of the same size as the input image. This helps the network to spread the watermark information in a wider space. Therefore, by removing some parts of the image by attacks, the watermark information can be retrieved from other parts. Using dilation layers is another innovation of the proposed method that helps the watermark spread in the wider neighborhood. This will improve the robustness of the method against image processing attacks during extraction. Instead of diffusing the watermark data into the input image, we inject the data into the feature space and force the network to do this in regions that increase the robustness against various attacks. Having embedding, extraction, and attack layer in an end-to-end network helps increase robustness and imperceptibility. Experimental results show the proposed method's superiority in imperceptibility and robustness compared to the state-of-the-art algorithms.

关键词： Blind watermarking convolutional auto-encoder Dilation layer Robustness Data expanding

来源：评论

学校读者我要写书评

暂无评论

automatic forensic identification using 3D sphenoid sinus segmentation and deep characterization

引用

MEDICAL & BIOLOGICAL ENGINEERING & COMPUTING 2020年第2期58卷 291-306页

作者： Souadih, Kamal Belaid, Ahror Ben Salem, Douraied Conze, Pierre-Henri Univ Abderrahmane Mira Med Comp Lab LIMED Bejaia 06000 Bejaia Algeria Lab Med Informat Proc LaTIM UMR 1101 Inserm 22 Ave Camille Desmoulins F-29238 Brest France CHRU Cavale Blanche Neuroradiol Dept Blvd Tanguy Prigent UBO F-29609 Brest France IMT Atlantique Technopole Brest Iroise F-29238 Brest France

Recent clinical research studies in forensic identification have highlighted the interest in sphenoid sinus anatomical characterization. Their pneumatization, well known as extremely variable in degrees and directions, could contribute to the radiologic identification, especially if dental records, fingerPrints, or DNA samples are not available. In this paper, we present a new approach for automatic person identification based on sphenoid sinus features extracted from computed tomography (CT) images of the skull. First, we present a new approach for fully automatic 3D reconstruction of the sphenoid hemisinuses which combines the fuzzy c-means method and mathematical morphology operations to detect and segment the object of interest. Second, deep shape features are extracted from both hemisinuses using a dilated residual version of a stacked convolutional auto-encoder. The obtained binary segmentation masks are thus hierarchically mapped into a compact and low-dimensional space preserving their semantic similarity. We finally employ the l(2) distance to recognize the sphenoid sinus and therefore identify the person. This novel sphenoid sinus recognition method obtained 100% of identification accuracy when applied on a dataset composed of 85 CT scans stemming from 72 individuals. automatic Forensic Identification using 3D Sphenoid Sinus Segmentation and Deep Characterization from Dilated Residual auto-encoders

关键词： Sphenoidal sinus convolutional auto-encoder 3D segmentation Fuzzy C-means Mathematical morphology Forensic identification

来源：评论

学校读者我要写书评

暂无评论

Multi-source data fusion for economic data analysis

引用

NEURAL COMPUTING & APPLICATIONS 2021年第10期33卷 4729-4739页

作者： Li, Menggang Wang, Fang Jia, Xiaojun Li, Wenrui Li, Ting Rui, Guangwei Beijing Jiaotong Univ Natl Acad Econ Secur Beijing Peoples R China Beijing Jiaotong Univ Beijing Lab Natl Econ Secur Early Warning Engn Beijing Peoples R China Beijing Jiaotong Univ Beijing Ctr Ind Secur & Dev Res Beijing Peoples R China Beijing Jiaotong Univ Sch Econ & Management Beijing Peoples R China

Economic data include data of various types and characteristics such as macro-data, meso-data, and micro-data. The source of economic data can be the data related to economy held by the National Bureau of statistics and a various software. These multi-source and heterogeneous data have important value for economic analysis and forecasting. Taking into account the limitations of existing methods such as low accuracy and complex calculations, this paper proposes an economic data analysis and prediction method based on machine learning. We use machine learning to solve the data fusion problem in the process of multi-source data analysis and prediction in the economic field. Specifically, we proposes an economic data analysis and forecasting method combining convolutional auto-encoder and extreme gradient boosting algorithms. This method uses a convolutional auto-encoder to extract the data characteristics of the normalized parameter sequence and uses it to train an extreme gradient boosting model to predict the level of economic development and evaluate the importance of each influencing factor. Finally, through a case study, this paper integrates the data of labor force, education and population to forecast GDP. Through the verification of this case, the prediction accuracy of the proposed method is higher than the AE-XGBoost method and CAE-1D-XGBoost method used in this experiment, and the error is kept below 11.7%.

关键词： Economic forecasts Multi-source data fusion convolutional auto-encoder Feature extraction Extreme gradient boosting

来源：评论

学校读者我要写书评

暂无评论

CSiamese: a novel semi-supervised anomaly detection framework for gas turbines via reconstruction similarity

引用

NEURAL COMPUTING & APPLICATIONS 2023年第22期35卷 16403-16427页

作者： Liu, Dan Zhong, Shisheng Lin, Lin Zhao, Minghang Fu, Xuyun Liu, Xueyun Harbin Inst Technol Sch Mechatron Engn Harbin 150006 Peoples R China Harbin Inst Technol Sch Ocean Engn Weihai 264209 Peoples R China Harbin Inst Technol Weihai Key Lab Intelligent Operat & Maintenance Weihai 264209 Peoples R China

The main problem in data-driven anomaly detection of gas turbines is that the monitoring data consists of only a very small number of abnormal samples with the overwhelming majority of normal samples. To address this problem, this paper develops a novel semi-supervised anomaly detection framework, namely CSiamese, and the parameters of the framework are optimized only using normal samples, with the ultimate purpose of improving the anomaly detection performance on imbalanced data sets. First, the convolutional auto-encoder is used to learn the reconstructed representation of the input sample. Second, the Siamese network is selected to learn to measure the similarity between the input and its reconstructed representation under noise conditions. Besides, a new loss function is developed by improving the contrastive loss, namely triangle loss, and can reduce the risk of collapsing solutions of the Siamese network when only using positive sample pairs. Third, maximum likelihood estimation is used to set the proper detection threshold to separate abnormal samples from normal samples. Finally, the effectiveness of the developed CSiamese has been evaluated using the real monitoring data of gas turbines and a public CIFAR-10 data set.

关键词： Anomaly detection convolutional auto-encoder The Siamese network Threshold selection

来源：评论

学校读者我要写书评

暂无评论

Deep semantic preserving hashing for large scale image retrieval

引用

MULTIMEDIA TOOLS AND APPLICATIONS 2019年第17期78卷 23831-23846页

作者： Zareapoor, Masoumeh Yang, Jie Jain, Deepak Kumar Shamsolmoali, Pourya Jain, Neha Kant, Surya Shanghai Jiao Tong Univ Inst Image Proc & Pattern Recognit Shanghai 200240 Peoples R China Chinese Acad Sci Inst Automat Beijing Peoples R China Jaypee Univ Engn & Technol Guna India India Inst Technol Roorkee Uttar Pradesh India

Hashing approaches have got a great attention because of its efficient performance for large-scale images. This paper, aims to propose a deep hashing method which can combines stacked convolutional autoencoder with hashing learning, where the input image hierarchically maps to the low dimensional space. The proposed method DCAH contains encoder-decoder, and supervisory sub-network, that generates a low dimensional binary code in a layer-wised manner of the deep conventional neural network. To optimizing the hash algorithm, we added some extra relaxations constraint to the objective function. In our extensive experiments on ultra-high dimensional image datasets, our results demonstrate that the decoder structure can improve the hashing method to preserve the similarities in hashing codes;also, DCAH achieves the best performance comparing to other states of the art approaches.

关键词： convolutional auto-encoder Image hashing Image retrieval Deep learning Similarity search Learning to hash

来源：评论

学校读者我要写书评

暂无评论

A denoising-classification neural network for power transformer protection

引用

Protection and Control of Modern Power Systems 2022年第1期7卷 801-814页

作者： Zongbo Li Zaibin Jiao Anyang He Nuo Xu School of Electrical Engineering Northeast Electric Power UniversityJilinJilinChina School of Electrical Engineering Xi’an Jiaotong UniversityXi’anShaanxiChina

Artificial intelligence(AI)can potentially improve the reliability of transformer protection by fusing multiple ***,owing to the data scarcity of inrush current and internal fault,the existing methods face the problem of poor *** this paper,a denoising-classification neural network(DCNN)is proposed,one which inte-grates a convolutional auto-encoder(CAE)and a convolutional neural network(CNN),and is used to develop a reli-able transformer protection scheme by identifying the exciting voltage-differential current curve(VICur).In the DCNN,CAE shares its encoder part with the CNN,where the CNN combines the encoder and a *** on the inter-action of the CAE reconstruction process and the CNN classification process,the CAE regards the saturated features of the VICur as noise and removes them ***,it guides CNN to focus on the unsaturated features of the *** unsaturated part of the VICur approximates an ellipse,and this significantly differentiates between a healthy and faulty ***,the unsaturated features extracted by the CNN help to decrease the data ergodicity requirement of AI and improve the ***,a CNN which is trained well by the DCNN is used to develop a protection *** simulations and dynamic model experiments verify its superior performance.

关键词： Transformer protection Exciting voltage-differential current curve convolutional auto-encoder convolutional neural network Denoising-classification neural network

来源：评论

学校读者我要写书评

暂无评论

Control chart recognition based on the parallel model of CNN and LSTM with GA optimization

引用

EXPERT SYSTEMS WITH APPLICATIONS 2021年 185卷 115689-115689页

作者： Yu, Yaoxiang Zhang, Min Southwest Jiaotong Univ Sch Mech Engn Chengdu 61003 Peoples R China Technol & Equipment Rail Transit Operat & Mainten Chengdu 610031 Peoples R China

Quality control process has become one of the most critical issues in intelligent manufacturing. As the most practical and prevalent tools for continuously monitoring, control chart patterns (CCPs) can be automatically recognized to judge the fault of production process. However, there are three shortcomings in the previous researches. 1) Insufficient or superfluous features are considered;2) Few studies simultaneously take the local features and time sequence information into account;3) With the exception of empiricism, little work has been done on searching for the optimal hyper-parameters for the neural networks. Accordingly, FFR-GACLN, a method for ten CCPs recognition is proposed in this paper. Such a method is comprised of two sections including the fusion feature reduction (FFR) and GACLN model construction. Firstly, the feature extraction comprising of features fusion and feature dimensionality reduction by convolutional auto-encoder is applied to enhance the reliability and performance of the extracted features. Then, such features are input into GACLN, a network combining the convolutional neural network (CNN) and the long-short term memory (LSTM) for achieving the final recognition task. Moreover, genetic algorithm (GA) is applied to search for the optimal hyper-parameters for FFR-GACLN. The quantitative and simulation results demonstrate that the performance of the proposed method is superior over the previous techniques.

关键词： Control chart patterns recognition Fusion feature reduction convolutional auto-encoder Long-short term memory convolutional neural network Genetic algorithm

来源：评论

学校读者我要写书评

暂无评论

Place perception from the fusion of different image representation

引用

PATTERN RECOGNITION 2021年 110卷 107680-107680页

作者： Li, Pei Li, Xinde Li, Xianghui Pan, Hong Khyam, M. O. Noor-A-Rahim, Md. Ge, Shuzhi Sam Southeast Univ Sch Automat Key Lab Measurement & Control CSE Minist Educ Nanjing Peoples R China Southeast Univ Sch Cyber Sci & Engn Nanjing Peoples R China Cent Queensland Univ Sch Engn & Technol Melbourne Vic Australia Univ Coll Cork Sch Comp Sci & IT Cork Ireland Natl Univ Singapore Interact Digital Media Inst Dept Elect & Comp Engn Singapore Singapore

Inspired by the human way of place understanding, we present a novel indoor place perception network to overcome: 1). the simplicity of existing methods that only use the image features of object regions to recognize the indoor place, 2). insufficient consideration of the semantic information about object at-tributes and states. By utilizing multi-modal information containing the image and natural language, the proposed method can comprehensively express the attributes, state, and relationships of objects which are beneficial for indoor place understanding and recognition. Specifically, we first present a natural language generation framework based on a Convolution Neural Network (CNN) and Long Short-Term Memory (LSTM) to imitate the process of place understanding. Next, a convolutional auto-encoder (CAE) and a mixed CNN-LSTM are proposed to extract image features and semantic features, respectively. Then, two different fusion strategies, namely feature-level fusion and object-level fusion, are designed to integrate different types of features and features from different objects. The category of the indoor place is finally recognized based on fused information. Comprehensive experiments are conducted on public datasets, and the results verify the effectiveness of the proposed place perception method based on linguistic cues. (c) 2020 Elsevier Ltd. All rights reserved.

关键词： Indoor place perception CNN LSTM convolutional auto-encoder Natural language

来源：评论

学校读者我要写书评

暂无评论

M³Net: A multi-model, multi-size, and multi-view deep neural network for brain magnetic resonance image segmentation

引用

PATTERN RECOGNITION 2019年 91卷 366-378页

作者： Wei, Jie Xia, Yong Zhang, Yanning Northwestern Polytech Univ Natl Engn Lab Integrated Aerospace Ground Ocean B Sch Comp Sci & Engn Xian 710072 Shaanxi Peoples R China Northwestern Polytech Univ Sch Comp Sci & Engn CMCC Xian 710072 Shaanxi Peoples R China

Segmentation of the brain into gray matter, white matter, and cerebrospinal fluid (CSF) using magnetic resonance (MR) imaging plays a fundamental role in neuroimaging research and clinical settings. Due to the complexity of brain anatomy, low image quality, and insufficient training data, both traditional and deep learning segmentation methods have a limited performance. In this paper, we propose a multi model, multi-size and multi-view deep neural network (M(3)Net) for brain MR image segmentation, which uses three identical modules to segment transaxial, coronal, and sagittal MR slices, respectively. Each module consists of multi-size U-Nets and multi-size back propagation neural networks. It also uses a probabilistic atlas to explore brain anatomy and a convolutional auto-encoder (CAE) to restore MR images. The proposed M(3)Net model was evaluated against widely used segmentation methods on both synthetic and clinical studies. Our results suggest that the proposed model is able to segment Brain MR Images more accurately. (C) 2019 Elsevier Ltd. All rights reserved.

关键词： Brain image segmentation Deep learning U-Net convolutional auto-encoder Back propagation neural network Magnetic resonance imaging image

来源：评论

学校读者我要写书评

暂无评论

Classification of electromyographic hand gesture signals using machine learning techniques

引用

NEUROCOMPUTING 2020年 401卷 236-248页

作者： Jia, Guangyu Lam, Hak-Keung Liao, Junkai Wang, Rong Kings Coll London Dept Engn London WC2R 2LS England

The electromyogram (EMG) signals from an individual's muscles can reflect the biomechanics of human movement. The accurate classification of individual and combined finger movements using surface EMG signals is able to support many applications such as dexterous prosthetic hand control. The existing research of EMG-based hand gesture classification faces the challenges of inaccurate classification, insufficient generalization ability and weak robustness. To address these problems, this paper proposes a deep learning model that combines convolutional auto-encoder and convolutional neural network (CAE+CNN) to classify an EMG dataset consisting of 10 classes of hand gestures. The proposed method shrinks the inputs into a smaller latent space representation using CAE and the resultant compressed features are served as inputs of CNN, which reduces the redundancy of EMG signals and improves the classification accuracy and training efficiency. Besides, to enhance the robustness and generalization ability for classification, a data processing approach is proposed which combines the windowing method and majority voting of the obtained results from the classifier. In addition, comprehensive comparative study is carried out with 8 widely applied and state-of-the-art classifiers in terms of classification accuracy, robustness subject to noise and statistical analysis (sensitivity, specificity, precision, F1 Score and Matthews correlation coefficient). The results demonstrates that the integration of windowing method, CAE+CNN and majority voting achieves the best performance (99.38% test accuracy for the data without adding noise, which is 3.78% higher than the best classifier used for comparison), strongest robustness (achieved 98.13% test accuracy when Gaussian noise of level le-5 is added to the raw dataset, which is 4.07% higher than the best classifier used for comparison) and statistical properties compared to other classifiers, which shows the potential for healthcare app

关键词： convolutional auto-encoder convolutional neural networks Deep learning EMG signals classification Machine learning

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：