检索结果-内蒙古大学图书馆

Abnormal event detection in surveillance videos through LSTM auto-encoding and local minima assistance

Discover Internet of Things 2025年第1期5卷 1-25页

作者： Sengonul, Erkan Samet, Refik Abu Al-Haija, Qasem Alqahtani, Ali Alsemmeari, Rayan A. Alghamdi, Bandar Alturki, Badraddin Alsulami, Abdulaziz A. Department of Computer Engineering Ankara University Ankara06100 Turkey Department of Cybersecurity Faculty of Computer & Information Technology Jordan University of Science and Technology PO Box 3030 Irbid22110 Jordan Department of Networks & Communications Engineering College of Computer Science & Information Systems Najran University Najran61441 Saudi Arabia Department of Information Technology Faculty of Computing and Information Technology King Abdulaziz University Jeddah21589 Saudi Arabia Department of Information Systems Faculty of Computing and Information Technology King Abdulaziz University Jeddah21589 Saudi Arabia

Abnormal event detection in video surveillance is critical for security, traffic management, and industrial monitoring applications. This paper introduces an innovative methodology for anomaly detection in video data, encompassing three primary stages: preprocessing, feature learning, and anomaly detection. We employ background subtraction and noise reduction during preprocessing to refine the data. The feature-learning stage involves training an LSTM autoencoder to capture the essential features of normal video sequences. For anomaly detection, we map video data to a lower-dimensional space (latent code) and compare it against the distribution of codes from normal sequences. We determine regularity scores and identify local minima points exceeding a specified threshold while scrutinizing shadows between adjacent maxima to confirm and pinpoint anomalies. When tested on the CUHK Avenue and UMN datasets, our methodology demonstrated performance with AUCs of 93.8% and 94.1%, respectively, outperforming several baseline models. Our results show the high precision of our method that can detect anomalies, highlighting its potential advantages that it can achieve for enhancing systems of surveillance. © The Author(s) 2025.

关键词： Noise abatement

来源：评论

学校读者我要写书评

暂无评论

An effective medical image fusion method utilizing moth-flame optimization and coupled neural P systems

引用

Neural Computing and Applications 2025年 1-41页

作者： Dinh, Phu-Hung Le, Thi-Hong-Ha Giang, Nguyen Long Faculty of Computer Science and Engineering Thuyloi University 175 Tay Son Dong Da Hanoi Viet Nam Hong Duc University 565 Quang Trung Dong Ve Thanh Hoa Viet Nam Institute of Information Technology Vietnam Academy of Science and Technology Hanoi Viet Nam Graduate University of Science and Technology Vietnam Academy of Science and Technology Hanoi Viet Nam

Medical image fusion is crucial for various applications, yet existing fusion models often encounter issues such as low quality, information loss, and insufficient contrast. This paper presents a novel approach to medical image fusion that directly addresses these challenges by enhancing brightness and contrast, improving overall image quality, and preserving essential structural features. Our method begins by decomposing input images into a base component and two detail components using a three-component decomposition (TCD) method with a truncated Huber filter (THF). The detail components are subsequently fused using an integrated approach combining the structure tensor saliency detection operator (STSDO), distance-weighted regional energy (DWRE), fast guided filter (FGF), and coupled neural P system (CNPS). The base components are then fused through an adaptive method guided by the moth-flame optimization (MFO) algorithm. To evaluate our approach, we used a dataset of 120 pairs of medical images and assessed performance with metrics such as QMLI, QCI, QAG, QAB/F, QP, QW, QE, and QCB. Experimental results demonstrate that the proposed model significantly enhances brightness and contrast and more effectively preserves information compared to recent methods. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

关键词： Image fusion

来源：评论

学校读者我要写书评

暂无评论

Liberated Arabic Handwritten Text Recognition using Convolutional Recurrent Neural Networks

Iraqi Journal for Computer Science and Mathematics

引用

Iraqi Journal for computer science and Mathematics 2025年第2期6卷 72-81页

作者： AlRababah, Ahmad AbdulQadir Aljahdali, Mohammed Khalid Al jahdali, Abdulrahim Abdulhamid AlGhanmi, Mohammed Saleh Al Barazanchi, Israa Ibraheem Department of Computer Science Faculty of Computing and Information Technology in Rabigh King Abdulaziz University Saudi Arabia College of Engineering University of Warith Al-Anbiyaa Karbala Iraq

Arabic script is exhibited in a cursive style, which is a departure from the norm in many common languages, and the shapes of letters are contingent on their positions within words. The form of the first letter is influenced by the subsequent letter, middle letters are shaped by both preceding and succeeding letters, and the shape of the final letter is determined by the preceding letter. Additionally, certain letters are found to have strikingly similar shapes, making Arabic text recognition a formidable challenge in computer vision. The challenge of detecting and recognizing Arabic handwritten text is addressed in this paper by proposing a novel system that integrates two Deep Learning models without the constraints of a predefined dictionary or language model. Object Detection techniques are employed in the first model to accurately identify lines, words, and punctuation marks, providing a robust foundation for text recognition. In the second model, a powerful recognition system is developed, trained on the IFN/ENIT dataset to predict character sequences from handwritten text images. The architecture is comprised of a convolutional neural network (CNN) with residual connections, followed by a Bi-directional Long Short-Term Memory (BLSTM) layer, and culminates in a fully connected layer. State-of-the-art results in unconstrained Arabic text recognition tasks are achieved by this approach. © 2025 The Author(s).

关键词： Long short-term memory

来源：评论

学校读者我要写书评

暂无评论

Eddy Detection of Lakes in the Qinghai–Tibet Plateau Based on Optical Remote Sensing and Deep Learning

引用

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 2025年 18卷 11996-12007页

作者： Bai, Bingxin Mu, Lixia Yang, Jie Huang, Baoxiang Ma, Chunyong Chen, Ge Ocean University of China Faculty of Information Science and Engineering Qingdao266100 China Laoshan Laboratory Qingdao266237 China Ocean University of China College of Marine Technology Qingdao266100 China Qingdao University College of Computer Science and Technology Qingdao266071 China

Lake eddies are dynamic phenomena prevalent in large lake systems, playing a critical role in affecting lake physics, nutrient transport, and ecological balance. Efficient and precise detection of these features is essential for advancing scientific understanding of lake dynamics and for improving environment management strategies. Despite their importance, the study of lake eddies remains under-explored, and current detection methods are insufficient. This study addresses this gap by proposing a lake eddy identification framework based on the YOLOv7 deep learning model, utilizing Landsat 8/9 satellite imagery acquired from April 2013 to December 2023. Qinghai Lake, Selin Co, and Nam Co on the Qinghai–Tibet Plateau were selected as the study areas. The method involves the development of an image enhancement process to optimize the visibility of lake eddies in remote sensing imagery, followed by the implementation of the YOLOv7 model for detection. The framework identified 912 cyclonic eddies and 102 anticyclonic eddies on the three aforementioned lakes, providing detailed information on their spatial distribution and morphology. This approach demonstrates significant potential for advancing the study of lake hydrodynamics and energy transport substances in aquatic ecosystems. © 2008-2012 IEEE.

关键词： Aquatic ecosystems

来源：评论

学校读者我要写书评

暂无评论

A novel approach to deriving adaboost classifier weights using squared loss function for overlapping speech detection

引用

Multimedia Tools and Applications 2025年 1-28页

作者： Asbai, Nassim Bounazou, Hadjer Zitouni, Sihem Department of Electronics Faculty of Sciences and Technology Mohamed El Bachir El Ibrahimi University Bordj Bou Arreridj City34030 Algeria Speech Com. & Signal Proc. Lab. Faculty of Electrical Engineering USTHB Bab Ezzouar City16111 Algeria Department of Computer Science Faculty of Mathematics and Computer Science Mohamed El Bachir El Ibrahimi University Bordj Bou Arreridj City34030 Algeria

In this paper, we address the complex problem of detecting overlapping speech segments, a key challenge in speech processing with applications in speaker diarization, automatic transcription, and multi-speaker recognition systems. Traditional approaches, often relying on exponential loss functions within the AdaBoost framework, struggle to maintain robustness in noisy or imbalanced data environments. To enhance detection accuracy, we propose a novel AdaBoost variant—SLFAdaBoost—utilizing a squared loss function specifically tailored for overlapping speech detection. This model demonstrates superior stability and convergence, significantly outperforming conventional methods. Experimental evaluation on the NIST 2005 corpus revealed that SLFAdaBoost achieves an accuracy range of 92.5% to 98.6%, with F1 scores between 95.4% and 99.2%, notably surpassing standard AdaBoost and Random Forest classifiers. Additionally, our model shows resilience in noisy conditions, maintaining precision and recall metrics at higher noise levels than other classifiers. These results underscore SLFAdaBoost's capacity to handle intricate data patterns, offering a more robust solution for real-world speech processing applications where overlapping segments are prevalent. This contribution provides an efficient, high-performing model, advancing the capabilities of ensemble methods in complex audio environments. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Speech recognition

来源：评论

学校读者我要写书评

暂无评论

Tree search hyper-heuristic with application to combinatorial optimization

引用

Journal of Heuristics 2025年第2期31卷 1-42页

作者： Gil-Gala, Francisco Javier Ɖurasevic, Marko Sierra, Maria R. Varela, Ramiro Department of Computer Science University of Oviedo Gijón 33213 Spain Faculty of Electrical Engineering and Computing University of Zagreb Zagreb 10000 Croatia

In this study, we investigate using the state space search paradigm to construct heuristics in the form of Priority Rules for combinatorial optimisation problems. This is an alternative to Genetic Programming (GP) and other hyper–heuristics, which represent the most common approach currently used. To do that, we define the problem of designing heuristics as a Constraint Satisfaction Problem and then exploit Any-Time Depth-First Search to solve it. To limit the effective size of the search space, we introduced a set of powerful pruning mechanisms, some embedded into the problem definition as constraints, while others by means of constraint propagation procedures. To further reduce the search space, we propose a heuristic procedure that allows the algorithm to discard some non-promising PRs, at low computational cost. The proposed approach, termed Systematic Search and Heuristic Evaluation (SSHE), was evaluated on two hard combinatorial optimisation problems, namely the One Machine Scheduling Problem with time-varying capacity (denoted by (1,Cap(t)||∑Tj)) and the classic Travelling Salesman Problem. The results of the experimental study show that SSHE is quite competitive with GP in building PRs;in particular, the PRs obtained by SSHE and GP showcase similar performance, but the ones produced by SSHE have generally lower size and so better readability than the PRs produced by GP. © The Author(s) 2025.

关键词： Heuristic search Hyper–heuristics One machine scheduling problem Priority rules Travelling salesman problem

来源：评论

学校读者我要写书评

暂无评论

Scalable Subspace Clustering via Purity Kernel Tensor Learning 3

Scalable Subspace Clustering via Purity Kernel Tensor Learni...

引用

3rd International Conference on Integrated Circuits and Communication Systems, ICICACS 2025

作者： Adhvaryu, Rachit Kannan, S. Rajes Sanjay, C.P. Dange, Pravin Parul University Department of Computer Science and Engineering Faculty of Engineering and Technology Gujarat India Chennai Institute of Technology Department of Computer Science and Engineering Kundrathur India Sri Balaji University Pune India Pune India

ISBN: (纸本)9798331508456

Subspace clustering is a fundamental task in machine learning, particularly for high-dimensional datasets where the goal is to group data points that lie in distinct subspaces. Traditional methods, such as Sparse Subspace Clustering (SSC) and Low-Rank Representation (LRR), have demonstrated good performance but often suffer from high computational costs, particularly when applied to large-scale data. To address these limitations, this paper introduces a novel technique called Purity Kernel Tensor Learning (PKTL). PKTL combines the power of tensor decomposition with a purity kernel function to improve both the efficiency and effectiveness of subspace clustering. By utilizing tensor-based representations, the method captures higher-order relationships in the data, allowing for a more robust clustering process. The purity kernel function measures the coherence within the clusters, enhancing the identification of subspaces in complex datasets. The proposed approach not only reduces computational complexity but also improves the clustering accuracy compared to existing methods. Experiments on synthetic and real-world datasets, including ORL, Yale, and MNIST, show that PKTL outperforms conventional subspace clustering algorithms, both in terms of clustering accuracy and scalability. The results demonstrate the potential of PKTL in handling large datasets while maintaining high performance, making it a promising approach for scalable subspace clustering in real-world applications. © 2025 IEEE.

关键词： Large datasets

来源：评论

学校读者我要写书评

暂无评论

DiffuSAR: Frequency Domain-Aware Diffusion Model for SAR Image Generation

引用

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 2025年 18卷 11851-11866页

作者： Ying, Zilu Ke, Wenyu Zhai, Yikui Long, Zhihao Zhou, Jianhong Zhu, Hufei Chen, C. L. Philip Wuyi University School of Electronics and Information Engineering Jiangmen529020 China South China University of Technology School of Electronics and Information Engineering Guangzhou510000 China South China University of Technology Faculty of Computer Science and Engineering Guangzhou510000 China

In recent years, deep learning-based synthetic aperture radar (SAR) image detection, recognition, and segmentation models achieve remarkable accuracy when trained on large amounts of SAR image samples. However, the acquisition of SAR images tends to be costly. A practical approach to address this issue involves the use of artificially generated training samples to supplement the dataset. Current SAR image generation methods based on generative adversarial networks have issues with generation fidelity and are challenge to converge. Moreover, most of these methods only consider the spatial domain of SAR images without exploring the characteristics of their frequency domain. To tackle the aforementioned problems, we proposed a lightweight frequency domain-aware SAR image generation model based on the denoising diffusion probabilistic model. The proposed generative model is capable of producing highly realistic artificial SAR image samples while converging stably. Meanwhile, our research reveals that high-frequency (HF) components of SAR images play a crucial role in the generation process, and a frequency adjustment module (FAM) based on the fast Fourier transform is implemented in our generation model to prevent interference of HF components by low-frequency components. Tests on the MSTAR dataset highlight our model's advantages in both generation quality and parameter efficiency compared to other cutting-edge image generation models. An additional ablation study also confirms the effectiveness of our proposed FAM, which reduces the Fréchet inception distance by 2.85%, confirming our hypothesis that HF components are more crucial to SAR image generation. © 2008-2012 IEEE.

关键词： Fast Fourier transforms

来源：评论

学校读者我要写书评

暂无评论

Enhancing structural health monitoring with AI-ML algorithms: a focus on crack detection and prediction

引用

Asian Journal of Civil engineering 2025年第5期26卷 1907-1918页

作者： Bader, Ahmad Shtayat, Amir Al-Mistarehi, Bara’ Communication and Computer Department Faculty of Engineering Jadara University Irbid Jordan Department of City Planning and Design Jordan University of Science and Technology Irbid 22110 Jordan Department of Civil Engineering Jordan University of Science and Technology Irbid Jordan

SHM is a very important process in terms of the safety and durability of infrastructure. Traditional SHM often faces problems detecting minor structural defects and handling large datasets. Therefore, certain more advanced approaches are called for. The paper discussed the applications of AI and ML algorithms, such as CatBoost and the African Vultures Optimization Algorithm, for such challenges. The research is based on a unique dataset of 8,541 rows and diverse features, developing a predictive framework that enhances crack detection and forecast capabilities. The approach mainly deals with heterogeneous data using the CatBoost algorithm, given its capability for high-accuracy predictions, while AVOA optimizes feature selection, reduces the computational cost, and guarantees no loss in model performance. This methodology has resulted in a significant enhancement of the prediction accuracy, which states the importance of AI-ML integration in SHM. The key results demonstrate the effectiveness of the model in detecting structural anomalies and crack propagation to enable proactive maintenance strategies. This study’s contributions have gone toward advancing SHM with scalable and efficient AI-ML frameworks, enabling real-time monitoring for better infrastructure management. Such development might have a transforming potential to cut down on maintenance costs and enhance operational safety, thus further encouraging sustainable infrastructure systems. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2025.

关键词： African vultures optimization algorithm CatBoost Crack Detection Feature selection Machine learning Structural health monitoring

来源：评论

学校读者我要写书评

暂无评论

Enhanced Small Lesion Detection on Grape Leaves Using Optimized YOLOv7 3

Enhanced Small Lesion Detection on Grape Leaves Using Optimi...

引用

3rd International Conference on Integrated Circuits and Communication Systems, ICICACS 2025

作者： Mahalakshmi, S. Gandhi, Jay Prabakaran, V. Rastogi, Shailesh Chennai Institute of Technology Department of Computer Science and Engineering Kundrathur Chennai India Faculty of Engineering and Technology Parul University Department of Computer science and Engineering Vadodara India Pune India Department of Finance Pune India

ISBN: (纸本)9798331508456

The detection of small lesions on grape leaves plays a pivotal role in the early identification and management of plant diseases, potentially reducing significant agricultural losses. Traditional detection systems, including advanced versions such as YOLOv7, often overlook small target diseases due to resolution constraints and inadequate feature extraction capabilities. Addressing this, our research enhances the YOLOv7 architecture by incorporating a specialized prediction branch and leveraging an improved channel attention mechanism in conjunction with an Extended-Efficient Long-range Attention Network (E-ELAN). Firstly, a new detection head specifically designed for small targets enhances the resolution capabilities of the system. Secondly, the integration of asymmetric convolution into the E-ELAN facilitates refined multi-scale feature extraction, critical for distinguishing minute pathological changes. Further, we enhance the model's sensitivity to small-scale features through a revamped channel attention mechanism. Moreover, we replace the conventional Complete Intersection over Union (CIoU) with a Structured Intersection over Union (SIoU) in the loss function, significantly improving the model's localization accuracy. Experimental results from testing the optimized YOLOv7 on a dataset of grape leaves affected by three prevalent diseases show an average accuracy improvement of 2.7%, achieving a remarkable 93.5%. This substantiates our model's efficacy in enhancing small lesion detection, thereby contributing effectively to early disease intervention strategies in viticulture. © 2025 IEEE.

关键词： asymmetric convolution channel attention mechanism plant disease diagnostics small lesion detection YOLOv7

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：