检索结果-内蒙古大学图书馆

Defocus deblur method of multi-scale depth-of-field cross-stage fusion image based on defocus map forecast

JOURNAL OF SUPERCOMPUTING 2025年第2期81卷 1-25页

作者： Li, Pei Bai, Tong Pan, Xiaoying Zuo, Chengyu Shaanxi Key Lab Network Data Anal & Intelligent Pr Xian Peoples R China Xian Key Lab Big Data & Intelligent Comp Xian Peoples R China Shaanxi Key Lab Intelligent Media Comp & Interact Xian Peoples R China

Defocus blur often arises in images captured with wide apertures and shallow depth of field, presenting significant challenges due to its spatial variability and difficulty in estimation. Existing methods for defocus and deblurring typically address overall blur but struggle with varying blur effects across different depths of field. To tackle this issue, we propose a novel approach that estimates defocus maps and performs multi-scale defocus deblurring through cross-stage fusion of multi-scale depth-of-field images. Our method, inspired by optimizing the pupil mask in monocular ranging, utilizes all-pixel dual-core focus sensing technology to estimate defocus blur for each pixel. We independently solve for the blur kernels of left and right parallax images, establishing a connection between depth of field and defocus in the form of a defocus map. This approach enables the recovery of fully focused images through multi-scale depth-of-field cross-stage fusion. By leveraging global information features from dual-pixel image sensing, our method significantly improves the handling of defocus blurring across different depth-of-field scales and outperforms recent defocus and deblurring methods in both quantitative and qualitative assessments.

关键词： Defocus blur Extended depth of field Dual-pixel sensors Deature fusion Image recovery

来源：评论

学校读者我要写书评

暂无评论

EFRNet: Edge feature refinement network for real-time semantic segmentation of driving scenes

引用

DIGITAL SIGNAL procESSING 2025年 156卷

作者： Hou, Zhiqiang Qu, Minjie Cheng, Minjie Ma, Sugang Wang, Yunchen Yang, Xiaobao Xian Univ Posts & Telecommun Sch Comp Sci & Technol Xian 710121 Peoples R China Key Lab Network Data Anal & Intelligent Proc Shaan Xian 710121 Peoples R China

In the semantic segmentation field, the dual-branch structure is a highly effective segmentation model. However, the frequent downsampling in the semantic branch reduces the accuracy of features expression with increasing network depth, resulting in suboptimal segmentation performance. To address the above issues, this paper proposes a real-time semantic segmentation network based on Edge Feature Refinement (Edge Feature Refinement network, EFRNet). A dual-branch structure is used in the encoder. To enhance the accuracy of deep features expression in the network, an edge refinement module (ERM) is designed in the dual-branch interaction stage to refine the features of the two branches and improve segmentation accuracy. In the decoder, a Bilateral Channel Attention (BCA) module is designed, which is used to extract detailed information and semantic information of features at different levels of the network, and gradually restore small target features. To capture multi- scale context information, we introduce a Multi-scale Context Aggregation Module (MCAM), which efficiently integrates multi-scale information in a parallel manner. The proposed algorithm has experimented on Cityscapes and CamVid datasets, and reaches 78.8% mIoU and 79.6% mIoU, with speeds of 81FPS and 115FPS, respectively. Experimental results show that the proposed algorithm effectively improves segmentation performance while maintaining a high segmentation speed.

关键词： Dual-branch structure Real-time semantic segmentation Edge refinement Detailed information Multi-scale context information

来源：评论

学校读者我要写书评

暂无评论

Remaining useful-life prediction of lithium battery based on neural-network ensemble via conditional variational autoencoder

引用

APPLIED INTELLIGENCE 2025年第1期55卷 1-15页

作者： Zhang, Hengshan Guo, Kaijie Chen, Yanping Sun, Jiaze Xian Univ Posts & Telecommun Sch Comp Sci & Technol Xian Peoples R China Shanxi Key Lab Network Data Anal & Intelligent Pro Xian Peoples R China Xian Key Lab Big Data & Intelligent Comp Xian Peoples R China

Ensemble learning using deep neural networks has become prevalent in predicting the Remaining Useful Life (RUL) of Lithium Batteries (LiBs). However, owing to the predominant linearity of ensemble learning, capturing nonlinear relationships among base learners remains a persistent challenge. This study presents an RUL-prediction method for LiBs based on a neural-network ensemble via a Conditional Variational Autoencoder (CVAE). The proposed method serves as a nonlinear ensemble learning method and promises enhanced prediction performance while maintaining ease of implementation. The methodology entails several key steps. First, data smoothing is conducted via local weighted linear regression. Subsequently, a preliminary linear-ensemble phase is executed through an attention mechanism, which filters out extraneous information in the features and bolsters the importance of valid features. Subsequently, a nonlinear ensemble is accomplished by utilizing the CVAE, with truth labels serving as conditions. Finally, the efficacy of the proposed method is substantiated through experimentation, demonstrating its superior performance compared to the candidate methods.

关键词： Remaining useful-life prediction Locally weighted linear regression Conditional variational autoencoder Attention mechanism

来源：评论

学校读者我要写书评

暂无评论

A large scale group decision making with expert guidance via discrete conditional variational autoencoder

引用

APPLIED INTELLIGENCE 2025年第6期55卷 1-15页

作者： Zhang, Hengshan He, Adong Sun, Jiaze Chen, Yanping Xian Univ Posts & Telecommun Sch Comp Sci & Technol Xian Peoples R China Shaanxi Key Lab Network Data Anal & Intelligent Pr Xian Peoples R China Xian Key Lab Big Data & Intelligent Comp Organizat Xian Peoples R China

In Large Scale Group Decision Making (LSGDM), the differences in decision-makers' professional backgrounds and attitudes often lead to high-quality decisions being overshadowed by numerous low-quality decisions, thus affecting the accuracy of the final decision. This study proposes a new decision-making method to address this challenge. First, a few experts are invited to make decisions as cluster centers, followed by obtaining decisions from a large number of ordinary decision-makers. The ordinary decisions are then generated and modified using a Discrete Conditional Variational Autoencoder (DCVAE) to enhance decision quality while maintaining consistency with expert decisions. Finally, the normalized prediction selection rate (NPSR) and the Borda Count consensus method are integrated to obtain the final result. Experimental results demonstrate the effectiveness of this method in improving the quality of LSGDM, providing a new solution to the coexistence of high- and low-quality decisions.

关键词： Consistency Large scale group decision making Discrete conditional variational auto-encoder Consensus reaching process

来源：评论

学校读者我要写书评

暂无评论

The VAE-FastGA anomaly detection model based on subspace and weakly correlated ultra-high-dimensional data

引用

JOURNAL OF SUPERCOMPUTING 2025年第3期81卷 1-32页

作者： Wan, Junhang Chen, Yanping Gao, Cong Xian Univ Posts & Telecommun Sch Comp Sci & Technol Xian 710121 Peoples R China Shaanxi Key Lab Network Data Anal & Intelligent Pr Xian 710121 Peoples R China Xian Key Lab Big Data & Intelligent Comp Xian 710121 Peoples R China

The rapid development of technology has led to a continuous increase in data dimensions, and the detection of high-dimensional data growing from tens of dimensions to millions of dimensions. As a result, the negative impact of the dimensional curse on anomaly detection has become increasingly remarkable. Although subspace research has been quite extensive, it has been difficult to extend to ultra-high-dimensional (UHD) datasets. To address this issue, this paper proposes the VAE-FastGA subspace search model, which is based on the idea of feature selection and aims to efficiently screen out anomalous subspaces from weakly correlated UHD data. The model integrates variational autoencoder (VAE) networks with accelerated genetic algorithms (FastGA), using VAE reconstruction to guide the search process of the genetic algorithm. An acceleration mechanism is designed to increase search efficiency by discarding invalid attributes during genetic operations and iteratively preserving anomalous subspaces. To mitigate the effects of sample sparsity, a probabilistic statistical model is introduced to optimize the anomalous subspaces. Finally, the idea of hyperdimensional spheres is integrated to improve the detection of marginal anomalies. The experimental sections organize precision experiments based on the KDD-CUP'99, MNIST, and UCSD datasets, perform subspace evaluations based on the MNIST dataset, and conduct scalability research based on incrementally generated datasets. The results show that the proposed algorithm exhibits higher AUC values, better subspace quality, and improved time efficiency in ultra-high dimensions.

关键词： Weakly correlated UHD data VAE-FastGA Feature selection Scalability

来源：评论

学校读者我要写书评

暂无评论

CDZL: a controllable diversity zero-shot image caption model using large language models

引用

SIGNAL IMAGE AND VIDEO procESSING 2025年第4期19卷 1-8页

作者： Zhao, Xin Kong, Weiwei Liu, Zongyao Wang, Menghao Li, Yiwen Xian Univ Posts & Telecommun Xian 710121 Shannxi Peoples R China Shaanxi Key Lab Network Data Anal & Intelligent Pr Xian 710121 Shannxi Peoples R China

Zero-shot image caption generation enables machines to produce descriptions without meticulously curated training data. Addressing the issues present in current zero-shot image caption models, such as slow speed and low-quality captions, this paper proposes a controllable and diverse zero-shot image caption generation model based on a large model (Controlled Diverse Zero-Shot Image Captioning Model, CDZL). The CDZL model does not update parameters;instead, it uses an iterative approach to merge the predicted distributions of target words from various models to generate diverse image captions. By adding control signals, it can produce image captions with controllability. Leveraging the knowledge of large language models, CDZL makes the generated captions more aligned with the images. During the iterative process, the model incorporates the Metropolis-Hastings sampling idea, reducing the number of iterations by rejecting samples with excessively low scores, thereby accelerating the generation speed. Experimental results show that our method outperforms current state-of-the-art (SOTA) methods in most evaluation metrics and exhibits faster generation speed.

关键词： Zero-shot Image caption Large language models Diversity Controllability

来源：评论

学校读者我要写书评

暂无评论

GADM: data augmentation using Generative Adversarial Diffusion Model for pulse-based disease identification

引用

BIOMEDICAL SIGNAL procESSING AND CONTROL 2025年 100卷

作者： Fan, Lin Chen, Tengye He, Lang Wang, Zhongmin Zhang, Rong Xian Univ Posts & Telecommun Sch Comp Sci & Technol Xian 710121 Shaanxi Peoples R China Shaanxi Key Lab Network Data Anal & Intelligent Pr Xian 710121 Shaanxi Peoples R China Xian Key Lab Big Data & Intelligent Comp Xian 710121 Shaanxi Peoples R China

Pulse is an important health indicator that includes physiological parameters like heart rate, rhythm, and vascular elasticity. While deep learning methods for pulse diagnosis have been widely used, they are often constrained by limited data, especially for rare diseases. data augmentation using diffusion models can address these issues, but their sampling process consumes a lot of time and is difficult to apply in practice. To address these challenges, we introduce a novel generative adversarial diffusion model (GADM) that significantly enhances data sampling efficiency and achieves superior results in data augmentation. Our model employs a timestep-dependent discriminator to model each stage of the diffusion process. By utilizing the adversarial loss of generative adversarial networks (GANs), GADM achieves tighter lower bounds on the log-likelihood compared to the original diffusion model, resulting in an implicit model that can generate high-quality samples faster. Experimental results demonstrate that the proposed GADM obtains better sample quality and is also 10 times faster compared to the denoised diffusion probability model (DDPM). Additionally, compared to the accuracy of 88.28% obtained from training on the original imbalanced dataset, training on the class-balanced dataset enhanced using GADM achieved an accuracy of 94.33% in the hypertension classification task.

关键词： Pulse data augmentation Diffusion model Generative adversarial network Medical data generation

来源：评论

学校读者我要写书评

暂无评论

Membrane computing for IoT task offloading: An efficient multi-objective constrained optimization framework

引用

APPLIED SOFT COMPUTING 2025年 169卷

作者： Tuo, Shouheng Huyan, Yihao Fan, Ting Zhao, Yong Xian Univ Posts & Telecommun Sch Comp Sci & Technol Xian 710121 Shaanxi Peoples R China Shaanxi Key Lab Network Data Anal & Intelligent Pr Xian 710121 Shaanxi Peoples R China Xian Key Lab Big Data & Intelligent Comp Xian 710121 Shaanxi Peoples R China

The computational tasks associated with Internet of Things (IoT) applications have become increasingly complex, with IoT devices (IoTDs) now being utilized in a multitude of contexts across people's daily lives. In view of the restricted computing resources and battery life of IoT devices, an increasing number of computing tasks are being handled by cloud servers. However, the restricted communication range of cloud servers gives rise to significant communication costs. Consequently, these complex computational tasks are transferred to edge servers situated in close proximity to IoT devices, with the objective of mitigating transmission delay and costs. To address the challenge of offloading decisions and obtain a multi-objective offloading decision that satisfies multiple constraints in an edge-cloud scenario, this study presents a framework of multi-objective constrained evolutionary algorithms based on membrane computing for solving constrained multi-objective computational offloading problems (MCMS-CMOEA). The framework draws inspiration from the independent parallelism observed within membrane computing membranes and the information exchange between them. Its objective is to minimize latency and energy consumption. The framework is structured into three phases, with the aim of exploring the solution space and increasing the diversity of the final solution set in a stepwise manner. In order to investigate the performance of the MCMS-CMOEA algorithm, a comparative analysis was conducted, in which the proposed algorithm was tested against nine state-of-the-art Constrained Multi-Objective Evolutionary Algorithms (CMOEAs) in three constrained multi-objective benchmark suites with different characteristics and challenges. Furthermore, the MCMS-CMOEA is compared with four offloading schemes on computational offloading problems of varying dimensions. The convergence and diversity of the algorithm are evaluated by two different comprehensive metrics: inverted generation

关键词： Edge computing Membrane computing (MC) Constrained multi-objective optimization Internet of things (IoT)

来源：评论

学校读者我要写书评

暂无评论

IFE-Net: Integrated feature enhancement network for image manipulation localization

引用

IMAGE AND VISION COMPUTING 2025年 153卷

作者： Su, Lichao Dai, Chenwei Yu, Hao Chen, Yun Fuzhou Univ Key Lab Network Comp & Intelligent Informat Proc Fuzhou Peoples R China Fuzhou Univ Coll Comp & Data Sci Fuzhou 350116 Peoples R China

Image tampering techniques can lead to distorted or misleading information, which in turn poses a threat in many areas, including social, legal and commercial. Numerous image tampering detection algorithms lose important low-level detail information when extracting deep features, reducing the accuracy and robustness of detection. In order to solve the problems of current methods, this paper proposes anew network called IFE-Net to detect three types of tampered images, namely copy-move, heterologous splicing and removal. Firstly, this paper constructs the noise stream using the attention mechanism CBAM to extract and optimize the noise features. The high-level features are extracted by the backbone network of RGB stream, and the FEASPP module is built for capturing and enhancing the features at different scales. In addition, in this paper, the initial features of RGB stream are additionally supervised so as to limit the detection area and reduce the false alarm. Finally, the final prediction results are obtained by fusing the noise features with the RGB features through the Dual Attention Mechanism (DAM) module. Extensive experimental results on multiple standard datasets show that IFE-Net can accurately locate the tampering region and effectively reduce false alarms, demonstrating superior performance.

关键词： Tampered localization Attention mechanism Edge supervision

来源：评论

学校读者我要写书评

暂无评论

Semi-supervised hybrid contrastive learning for PolSAR image classification

引用

KNOWLEDGE-BASED SYSTEMS 2025年 311卷

作者： Hua, Wenqiang Sun, Nan Liu, Lin Ding, Chen Dong, Yizhuo Sun, Wei Xian Univ Posts & Telecommun Shaanxi Key Lab Network Data Anal & Intelligent Pr Xian 710121 Peoples R China Sch Comp Sci & Technol Xian Key Lab Big Data & Intelligent Comp Xian 710121 Peoples R China Chinese Acad Surveying & Mapping Beijing 100036 Peoples R China

Deep learning methods have been widely applied in polarimetric synthetic aperture radar (PolSAR) image classification. However, these methods require abundant labeled data to achieve satisfactory performance, and it requires considerable effort to obtain a good deal of labeled data in practice, which requires huge human and resources. To tackle these issues, a new semi-supervised hybrid contrastive learning (SSHCL) is proposed for PolSAR image classification of finite labeled data. First, a hybrid contrastive learning model is constructed to extract feature information using only a handful of labeled data and unlabeled data. Second, combined with the idea of semi-supervised learning, a semi-supervised samples selection method is proposed. The pseudo-labeled samples are selected from unlabeled samples by combining superpixel and contrastive similarity, and are included in the label dataset. Then, the extended labeled samples are used to train the network for a new round. Finally, the optimized SSHCL network is used to classify the pixels of the whole image. The experimental outcomes conducted on three real PolSAR datasets demonstrate that the superior performance of the proposed compared to other existing methods under finite labeled data.

关键词： Contrastive learning Polarimetric synthetic aperture radar Semi-supervised learning Image classification

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：