检索结果-内蒙古大学图书馆

DSTIGCN: Deformable Spatial-Temporal Interaction graph convolution Network for Pedestrian Trajectory Prediction

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 2025年第5期26卷 6923-6935页

作者： Chen, Wangxing Sang, Haifeng Wang, Jinyu Zhao, Zishan Shenyang Univ Technol Sch Informat Sci & Engn Shenyang 110870 Peoples R China

Accurate and reliable pedestrian trajectory prediction can reduce the risk of human-vehicle collisions and predict accidents in advance, which is crucial for developing autonomous driving and intelligent monitoring. Previous trajectory prediction methods face two common problems: 1. ignoring the joint modeling of pedestrians' complex spatial-temporal interactions, and 2. suffering from the long-tail effect, which prevents accurate capture of the diversity of pedestrians' future movements. To address these problems, we propose a Deformable Spatial-Temporal Interaction graph convolution Network (DSTIGCN). First, we construct a spatial graph and employ the attention mechanism to preliminarily describe the spatial interactions of pedestrians at each moment. To solve problem 1, we design a deformable spatial-temporal interaction module. The module autonomously learns the spatial-temporal interaction relationships of pedestrians through the offset of multiple asymmetric deformable convolution kernels in both spatial and temporal dimensions, thereby achieving joint modeling of complex spatial-temporal interactions. Next, we obtain trajectory representation features through graph convolution and then predict the two-dimensional Gaussian distribution parameters of future trajectories using the Temporal Attention-Gated Temporal convolution Network (TAG-TCN). To address problem 2, we introduce Latin hypercube sampling to sample the two-dimensional Gaussian distribution of future trajectories, thereby improving the multi-modal prediction effect of the model under limited samples. Experiments on ETH, UCY, and SDD datasets have verified that our method can achieve high-precision prediction of pedestrian future trajectories under limited parameters.

关键词： Pedestrian trajectory prediction graph convolution deformable convolution temporal convolution network latin hypercube sampling Pedestrian trajectory prediction graph convolution deformable convolution temporal convolution network latin hypercube sampling

来源：评论

学校读者我要写书评

暂无评论

Dynamic networks link prediction based on continuous gated recurrent graph convolution

引用

INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS 2025年第4期16卷 2653-2669页

作者： Liao, Yunchun Shu, Jian Liu, Linlan Nanchang Hangkong Univ Sch Software Nanchang 330063 Peoples R China Nanchang Hangkong Univ Sch Informat Engn Nanchang 330063 Peoples R China

Link prediction, as a fundamental task in dynamic networks, holds great developmental significance. Dynamic networks contain rich spatiotemporal features which are crucial for this task. However, existing deep learning models for learning spatiotemporal information in dynamic networks have shown certain limitations. Notably, they often fail to fully leverage the attributes inherent in dynamic networks and overlook local fine-grained evolutionary features. To address these shortcomings, a dynamic network link prediction method based on Continuous Gated Recurrent graph convolution Network is proposed. The method constructs initial node features by combining common neighbor indices with link attributes to better capture the network's inherent information. Specifically, we integrate Neural Ordinary Differential Equations to enhance the T-GCN model, creating a continuous gated recurrent graph convolution model that excels at capturing fine-grained evolutionary features. Then, a global evolution information extraction module is designed to learn the long and short-term spatiotemporal features comprehensively. Extensive experiments on real dynamic network datasets demonstrate that the proposed method outperforms existing baseline models.

关键词： Dynamic networks Link prediction graph convolution Neural ordinary differential equations

来源：评论

学校读者我要写书评

暂无评论

Reinforcement tokenization and graph convolution for high-precision breast tumor segmentation in DCE-MRI

引用

BIOMEDICAL SIGNAL PROCESSING AND CONTROL 2025年 100卷

作者： Akbar, Arslan Han, Suya Rehman, Naveed Urr Irshad, Rabia Ahmed, Kanwal Ali, Muhammad Mumtaz Al Mazroa, Alanoud Zhengzhou Univ Sch Elect & Informat Engn Zhengzhou 450001 Peoples R China Zhengzhou Univ Sch Comp & Artificial Intelligence Zhengzhou 450001 Peoples R China Beijing Univ Chem Technol Sch Chem Beijing 100013 Peoples R China Henan Univ Sch Software Kaifeng 475001 Peoples R China Princess Nourah bint Abdulrahman Univ Coll Comp & Informat Sci Dept Informat Syst Riyadh 11671 Saudi Arabia

Magnetic resonance imaging is essential for the surveillance and assessment of therapeutic efficacy in individuals predisposed to a heightened risk of breast cancer. The partitioning of malignant areas is a crucial stage in the thorough examination of breast MRI. However, achieving automatic and reliable segmentation remains highly tough due to the significant variations in cancer size, shape, and intensity appearance. Nevertheless, the majority of current techniques prioritize the extraction and selection of profound semantic characteristics, disregarding certain parameters that possess distinct meaning and significance for the segmentation task. This research presents a breast tumor segmentation technique that combines contextual mapping with advanced edge analysis from Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) scans. The objective is to obtain precise and accurate segmentation. The suggested technique primarily comprises a Contextual mapping module, an Edge analysis module, and a Feature integration module. The Contextual mapping module utilizes the Swin Transformer to extract contextual features and incorporates a shifting patch tokenization technique to enhance guidance. The edge analysis module is constructed using convolutional Neural Networks (CNNs), and a novel Edge-Aware Attention Module (EAAM) is introduced to enhance the image features. The aim of the feature integration module is to amalgamate the contextual features and the augmented edge details that have been delineated. In accomplishing this, we have employed graph convolution, which aids in feature reasoning and the dissemination of information, thereby facilitating efficient feature fusion. The proposed approach has been validated using two benchmark datasets. Our proposed model has demonstrated superior performance compared to both baseline models and state-of-the-art models, as evidenced by its higher scores in the Dice coefficient, Jaccard index, Precision, Sensitivity, and

关键词： Segmentation Breast tumor Deep learning graph convolution Reinforcement tokenization

来源：评论

学校读者我要写书评

暂无评论

Maxillofacial bone movements-aware dual graph convolution approach for postoperative facial appearance prediction

引用

MEDICAL IMAGE ANALYSIS 2025年 99卷 103350页

作者： Huang, Xinrui He, Dongming Li, Zhenming Zhang, Xiaofan Wang, Xudong Shanghai Jiao Tong Univ Sch Elect Informat & Elect Engn Shanghai 200240 Peoples R China Shanghai Jiao Tong Univ Shanghai Peoples Hosp 9 Sch Med Dept Oral Craniomaxillofacial Shanghai 200011 Peoples R China Shanghai Jiao Tong Univ Coll Stomatol Shanghai 200125 Peoples R China Natl Ctr Stomatol Shanghai 200011 Peoples R China Natl Clin Med Res Ctr Oral Dis Shanghai 200011 Peoples R China Shanghai Key Lab Stomatol Shanghai 200011 Peoples R China Shanghai Res Inst Stomatol Shanghai 200011 Peoples R China Chinese Acad Med Sci Res Unit Oral & Maxillofacial Regenerat Med Shanghai 200011 Peoples R China Shanghai Artificial Intelligence Lab Shanghai 200030 Peoples R China

Postoperative facial appearance prediction is vital for surgeons to make orthognathic surgical plans and communicate with patients. Conventional biomechanical prediction methods require heavy computations and time-consuming manual operations which hamper their clinical practice. Deep learning based methods have shown the potential to improve computational efficiency and achieve comparable accuracy. However, existing deep learning based methods only learn facial features from facial point clouds and process regional points independently, which has constrains in perceiving facial surface details and topology. In addition, they predict postoperative displacements for all facial points in one step, which is vulnerable to weakly supervised training and easy to produce distorted predictions. To alleviate these limitations, we propose a novel dual graph convolution based postoperative facial appearance prediction model which considers the surface geometry by learning on two graphs constructed from the facial mesh in the Euclidean and geodesic spaces, and transfers the bone movements to facial movements in dual spaces. We further adopt a coarse-to-fine strategy which performs coarse predictions for facial meshes with fewer vertices and then adds more to obtain more robust fine predictions. Experiments on real clinical data demonstrate that our method outperforms state-of-the-art deep learning based methods qualitatively and quantitatively.

关键词： Orthognathic surgery Facial appearance prediction Geometric deep learning graph convolution

来源：评论

学校读者我要写书评

暂无评论

TIGC-Net: Transformer-Improved graph convolution Network for spatio-temporal prediction

引用

BIOMEDICAL SIGNAL PROCESSING AND CONTROL 2025年 100卷

作者： Chen, Kai Zhou, Zhengyuan Liu, Yao Ji, Tianjiao Sun, Weiya Yang, Chunfeng Chen, Yang Lu, Xiao Southeast Univ Sch Cyber Sci & Engn Nanjing 210096 Peoples R China Southeast Univ Key Lab New Generat Artificial Intelligence Techno Minist Educ Nanjing 210096 Peoples R China Southeast Univ Coll Software Engn Nanjing 210096 Peoples R China Southeast Univ Sch Comp Sci & Engn Lab Image Sci & Technol Nanjing 210096 Peoples R China Southeast Univ Zhongda Hosp Dept Radiol Jiangsu Key Lab Mol & Funct Imaging Nanjing 210009 Peoples R China Southeast Univ Sch Comp Sci & Engn Jiangsu Prov Joint Int Res Lab Med Informat Proc Nanjing 210096 Peoples R China Chinese Ctr Dis Control & Prevent Natl Inst Viral Dis Control & Prevent NHC Key Lab Med Virol & Viral Dis Beijing Peoples R China Beijing Inst Tracking & Commun Technol Beijing 100094 Peoples R China Nanjing Med Univ Dept Rehabil Med Affiliated Hosp 1 300 Guangzhou Rd Nanjing 210029 Peoples R China

Modeling spatio-temporal sequences is an important topic yet challenging for existing neural networks. Most of the current spatio-temporal sequence prediction methods usually capture features separately in temporal and spatial dimensions or employ multiple mutually independent local spatio-temporal graphs to represent a spatio-temporal sequence. The first kind of method mentioned above is difficult to mine the complex spatiotemporal correlations, while the other is limited for the accuracy of long-term predictions. To handle these issues, this paper proposes a Transformer-Improved graph convolution Network for spatio-temporal prediction. Specifically, the temporal location encoding method is exploited to derive the spatio-temporal characteristics of the sequence utilizing a spatio-temporal feature fusion network. In addition, a spatio-temporal attention network is developed to enhance the spatio-temporal correlation of the sequence, and the dynamic spatial features of sequence are further extracted through the adaptive graph convolution network. A private dataset and a public dataset are employed to demonstrate the performance of the proposed TIGC-Net. The qualitative and quantitative results show that the proposed TIGC-Net can extract dynamic spatio-temporal properties more effectively, enhance the spatio-temporal correlation of sequences and improve the prediction accuracy compared with four state-of-the-art.

关键词： Spatio-temporal sequence prediction graph convolution Attention mechanism

来源：评论

学校读者我要写书评

暂无评论

ETGC2-net: an enhanced transformer and graph convolution combined network for landslide detection

引用

NATURAL HAZARDS 2025年第1期121卷 135-160页

作者： Fan, Shuaishuai Fu, Yibin Li, Weiming Bai, Hongyang Jiang, Yonghua Shandong Technol & Business Univ SDTBU Sch Informat & Elect Engn Yantai 264000 Shandong Peoples R China Nanjing Univ Sci & Technol NJUST Sch Energy & Power Engn Nanjing 210094 Jiangsu Peoples R China Wuhan Univ WHU Sch Remote Sensing & Informat Engn Wuhan 430072 Hubei Peoples R China

Landslide detection is one of the crucial tasks in geological hazard prevention and control. Accurate detection and prediction of landslide areas contribute to taking appropriate preventive and mitigation measures in advance. However, it still faces challenges in terms of timeliness and accuracy. Our design motivation is to seek a network that can demonstrate excellent performance and relatively low computational cost in practical landslide area detection tasks. The proposed method in this paper can simultaneously capture local information and global dependencies in the images while achieving outstanding performance with fewer parameters. Moreover, considering the spatial correlation of landslides in remote sensing images, we introduce graph convolution to model the internal topological relationship within areas, with superpixels serving as guidance. Extensive experiments on two publicly available landslide datasets demonstrate that our network outperforms traditional and existing state-of-the-art methods while achieving the optimal balance between efficiency and performance, thus exhibiting broad potential for practical applications.

关键词： Landslide detection Lightweight architecture Enhanced transformer graph convolution Superpixel segmentation

来源：评论

学校读者我要写书评

暂无评论

Transient Stability Preventive Control Based on graph convolution Neural Network and Transfer Deep Reinforcement Learning

引用

CSEE Journal of Power and Energy Systems 2025年第1期11卷 136-149页

作者： Tianjing Wang Yong Tang School of Electrical and Electronic Engineering Nanyang Technological UniversitySingapore Laboratory of Power Grid Safety and Energy Conservation China Electric Power Research InstituteBeijing100192China

This study proposes a new transient-stability preventive control (TSPC) method based on graph convolutional neural networks (GCNN) and transfer deep reinforcement learning (DRL) to address non-convergence problems of traditional optimization algorithms and slow training speed of artificial intelligence algorithms for TSPC. First, a transient stability assessor (TSA) with GCNN is developed to assess current-power-flow state. Sensitivities of the transient-stability index relative to the generators are approximately calculated using TSA;generators with significant influence able to narrow action space are identified. Subsequently, the Markov decision-making process of TSPC is derived by introducing the process of TSPC. A DRL for TSPC is constructed by adding entropy to twin delayed deep deterministic policy gradient (TD3). Knowledge learned by TSA is transferred to DRL based on transfer learning, which improves learning efficiency. Finally, case studies based on the IEEE 39-bus system and an actual power grid prove the effectiveness of the proposed method. Comparisons performed with reference algorithms in literature demonstrate the proposed method has better performance in both control effect and speed.

关键词： Deep reinforcement learning graph convolution preventive control transfer transient stability

来源：评论

学校读者我要写书评

暂无评论

Dual-Domain Self-Supervised Deep Learning with graph convolution for Low-Dose Computed Tomography Reconstruction

引用

JOURNAL OF IMAGING INFORMATICS IN MEDICINE 2025年 1-17页

作者： Yang, Feng Zhao, Feixiang Liu, Yanhua Liu, Min Liu, Mingzhe Chengdu Univ Technol Coll Nucl Technol & Automat Engn 1 East 3rd RdErxianqiao Chengdu 610059 Sichuan Peoples R China Univ Elect Sci & Technol China Sichuan Canc Hosp & Inst Sichuan Canc Ctr Dept Radiat OncolAffiliated Canc Hosp 55 Sect 4Renmin South Rd Chengdu 610054 Sichuan Peoples R China Chengdu Univ Technol Sch Mech & Elect Engn 1 East 3rd RdErxianqiao Chengdu 610059 Sichuan Peoples R China Wenzhou Univ Technol Sch Data Sci & Artificial Intelligence Wenzhou 325000 Zhejiang Peoples R China

X-ray computed tomography (CT) is a commonly used imaging modality in clinical practice. Recent years have seen increasing public concern regarding the ionizing radiation from CT. Low-dose CT (LDCT) has been proven to be effective in reducing patients' radiation exposure, but it results in CT images with low signal-to-noise ratio (SNR), failing to meet the image quality required for diagnosis. To enhance the SNR of LDCT images, numerous denoising strategies based on deep learning have been introduced, leading to notable advancements. Despite these advancements, most methods have relied on a supervised training paradigm. The challenge in acquiring aligned pairs of low-dose and normal-dose images in a clinical setting has limited their applicability. Recently, some self-supervised deep learning methods have enabled denoising using only noisy samples. However, these techniques are based on overly simplistic assumptions about noise and focus solely on CT sinogram denoising or image denoising, compromising their effectiveness. To address this, we introduce the Dual-Domain Self-supervised framework, termed DDoS, to accomplish effective LDCT denoising and reconstruction. The framework includes denoising in the sinogram domain, filtered back-projection reconstruction, and denoising in the image domain. By identifying the statistical characteristics of sinogram noise and CT image noise, we develop sinogram-denoising and CT image-denoising networks that are fully adapted to these characteristics. Both networks utilize a unified hybrid architecture that combines graph convolution and incorporates multiple channel attention modules, facilitating the extraction of local and non-local multi-scale features. Comprehensive experiments on two large-scale LDCT datasets demonstrate the superiority of DDoS framework over existing state-of-the-art methods.

关键词： Low-dose computed tomography Reconstruction Denoising Self-supervised deep learning graph convolution

来源：评论

学校读者我要写书评

暂无评论

TransgraphNet: A novel network for medical image segmentation based on transformer and graph convolution

引用

BIOMEDICAL SIGNAL PROCESSING AND CONTROL 2025年 104卷

作者： Zhang, Ju Ye, Zhiyi Chen, Mingyang Yu, Jiahao Cheng, Yun Hangzhou Normal Univ Coll Informat Sci & Technol Hangzhou 310030 Peoples R China Zhejiang Hosp Dept Med Imaging Hangzhou 310058 Peoples R China

Deep learning-based medical image segmentation has received great attention both from academic researchers and physicians and made significant progress in recent years. However, there is still some room to further improve the accuracy of segmentation. In this manuscript, a network, (TGNet)TransgraphNet, based on Transformer and graph convolution is proposed. The network exploits both the advantages of Transformers and graph convolutional networks (GCN). The proposed network has an encoder-decoder structure and learns effectively the global and local features simultaneously. With Transformer the encoder extracts global features, and through GCN, the decoder restores the spatial structure of the image, and thus the proposed network can understand the medical image more comprehensively and improve the segmentation accuracy. A spatial and channel parallel module (SCPM) is proposed, which is more flexible and can adjust the attention to image features at multiple levels. A fully convolutional Transformer attention module (FTAM) is developed to address local details. Together with SCPM, the FTAM module can understand the fine structures and features in the image, and further improve segmentation performances. With a graph convolutional network (GCN), a graph convolutional hybrid attention module (GCHA) is proposed to deal with irregular structures and global relationships in medical image segmentation. Extensive comparison experiments are conducted and the performances are evaluated on ACDC, Synapse, and Coronacases datasets, showing that the proposed network has better accuracy than most existing models. In particular, TransgraphNet achieves a performance improvement of 0.55% on the ACDC dataset, a 1.6% performance improvement on the Synapse dataset, and a 2.34% performance improvement on the Coronacases dataset. Ablation studies show that the proposed SCPM, FTAM, and GCHA modules improve the segmentation performances significantly.

关键词： Medical image segmentation Deep learning Transformer graph convolution

来源：评论

学校读者我要写书评

暂无评论

State of charge prediction for lithium-ion batteries based on multi-process scale encoding and adaptive graph convolution

引用

JOURNAL OF ENERGY STORAGE 2025年 113卷

作者： Wang, Hongyan Wei, Wu Cui, Langfu Zhou, Gongcheng Xiang, Gang Zhang, Qingzhen Beihang Univ Sch Automat Sci & Elect Engn Beijing 100191 Peoples R China Beijing Aerosp Automat Control Inst Beijing 100854 Peoples R China

Accurate prediction of the state of charge (SOC) of lithium-ion batteries in spacecraft power systems is crucial for optimizing energy management, effectively responding to emergencies, and enhancing system reliability and safety. To capture the slow degradation of chemical characteristics across multiple charge-discharge processes and cycles in lithium-ion batteries, this paper proposes a SOC prediction method based on multi-process scale encoding and adaptive graph convolution (MPSEAGC). This method employs adaptive graph convolution to capture the similarities and gradual differences between various charge-discharge processes of the lithium-ion battery and uses a multi-head attention mechanism to identify intra-sequence correlations within each charge or discharge process. By conducting comparative experiments using publicly available lithium-ion battery datasets, the proposed method achieved a mean squared error (MSE) of 0.062 % and a mean absolute error (MAE) of 1.652 %, representing improvements of 54.74 % and 42.26 % over existing models, particularly excelling in long-sequence prediction. This method provides an accurate and reliable approach to predicting the SOC of batteries, offering a promising solution for predicting battery performance and state in practical applications.

关键词： Lithium-ion battery State of charge prediction Deep learning network graph convolution Multi-head attention mechanism

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：