检索结果-内蒙古大学图书馆

2025 IEEE International conference on Acoustics, Speech, and signal processing, ICASSP 2025

作者： Yuan, Yunlong Guo, Yuanfan Wang, Chunwei Xu, Hang Zhang, Li School of Data Science Fudan University Shanghai China Noah's Ark Lab Huawei Shanghai China

ISBN: (纸本)9798350368741

Recent advances in diffusion models have greatly improved text-driven video generation. However, training models for long video generation demands significant computational power and extensive data, leading most video diffusion models to be limited to a small number of frames. Existing training-free methods that attempt to generate long videos using pre-trained short video diffusion models often struggle with issues such as insufficient motion dynamics and degraded video fidelity. In this paper, we present Brick-Diffusion, a novel, training-free approach capable of generating long videos of arbitrary length. Our method introduces a brick-to-wall denoising strategy, where the latent is denoised in segments, with a stride applied in subsequent iterations. This process mimics the construction of a staggered brick wall, where each brick represents a denoised segment, enabling communication between frames and improving overall video quality. Through quantitative and qualitative evaluations, we demonstrate that Brick-Diffusion outperforms existing baseline methods in generating high-fidelity videos. © 2025 IEEE.

关键词： brick-to-wall denoising Diffusion models long video generation training-free

来源：评论

学校读者我要写书评

暂无评论

Learning to Follow Infrared Prior Repersentation for Image Dehazing

Learning to Follow Infrared Prior Repersentation for Image D...

引用

2025 IEEE International conference on Acoustics, Speech, and signal processing, ICASSP 2025

作者： Du, Yiquan Yang, Ming Huo, Haiyang Shi, Jiaqi Ding, Jiangang Pei, Lili School of Information Engineering Chang'an University Shaanxi Xi'an710064 China School of Data Science and Artificial Intelligence Chang'an University Shaanxi Xi'an710064 China

ISBN: (纸本)9798350368741

The infrared image can distinguish the targets from the background based on radiation differences, providing more significant target visibility under dense haze. Fusion of visible haze images with infrared prior representations can generate high-quality fused images for high-level tasks. Consequently, we propose a novel dual-modal fusion network structure that makes full use of infrared prior representations for dehazing. Specifically, we emphasize a Multi-modal Feature Extraction Network (MMFE) to extract deep multi-scale features. Meanwhile, we introduce a Multi-scale Feature Extraction module (MSFE), integrating an Efficient Dual Attenion block (EDAB) to efficiently explore more spatial and marginal information. Additionally, we propose a new feature fusion strategy, which calculates feature fusion weights based on an adaptive multi-head self-attention. Therefore, IPRDehazeNet achieves better dehazing results through dual-modal fusion. Experimental results indicate that IPRDehazeNet outperforms various advanced methods. © 2025 IEEE.

关键词： Image dehazing Multi-modal Multi-scale Visible-Infrared fusion

来源：评论

学校读者我要写书评

暂无评论

Diffraction of a Nonharmonic signal on a Spatially Periodic Structure of a Sawtooth Profile

Diffraction of a Nonharmonic Signal on a Spatially Periodic ...

引用

2025 Systems of signals Generating and processing in the Field of on Board Communications, SOSG 2025

作者： Kryachko, A.F. Tyurina, A.I. State University of Aerospace Instrumentation St. Petersburg Russia

ISBN: (纸本)9798331532635

One of the directions of improving the characteristics of radar systems is the use of ultra-wideband sounding signals. This increases the radar's resolution in all coordinates, stealth of operation, noise immunity, and the possibility of identifying targets on the basis of their radio image. The probability of target detection and stability of target tracking increases. There are a number of difficulties in the process of generation, emission, propagation and reception of ultra-wideband signals. As a model of periodic inhomogeneities of the underlying surface in the propagation of non-harmonic signal in some cases use spatially periodic structures of sawtooth profile. The purpose of the research is to develop a numerical method for solving the problem of diffraction of non-sinusoidal electromagnetic waves on a spatially periodic structure of a sawtooth profile in order to model the propagation of ultrawideband signal over the sea surface. A semi-reversal and frequency method has been used, allowing the use of previously obtained solutions for the harmonic case. A modified method for calculating the scattered field when irradiating a sawtooth profile grating with an ultra-wideband electromagnetic wave is presented. The data of the experiment carried out on the ultrawideband measuring complex show a satisfactory coincidence with the results of numerical calculations and thus confirm the correctness of the mathematical model. The proposed numerical method is an effective means of taking into account the influence of periodic inhomogeneities of the underlying surface on the propagation of a non-harmonic electromagnetic wave. The influence of geometrical and electrophysical characteristics (period, angle of asymmetry, impedance) of the grating on the structure of the scattered field is investigated in detail. © 2025 IEEE.

关键词： Numerical methods

来源：评论

学校读者我要写书评

暂无评论

Ultrasound-Guided Registration Pseudo-Labels for Semi-Supervised Brachial Plexus Segmentation

Ultrasound-Guided Registration Pseudo-Labels for Semi-Superv...

引用

2025 IEEE International conference on Acoustics, Speech, and signal processing, ICASSP 2025

作者： Luo, Jia Zhang, Ze Ding, Yi Wang, Yiqian Zhang, Jian Network and Data Security Key Laboratory of Sichuan Province School of Information and Software Engineering University of Electronic Science and Technology of China Chengdu China Yibin GREAT Technology Co. Ltd. Yibin644000 China Sichuan Provincial Hospital for Women and Children/Affilated Women and Children's Hospital of Chengdu Medical College Chengdu610054 China

ISBN: (纸本)9798350368741

In semi-supervised medical image segmentation, two main challenges arise. First, the quality of pseudo-labels generated by segmentation networks in data-limited scenarios is often poor, reducing segmentation accuracy. Second, many methods fail to effectively utilize the temporal context in video data. Moreover, the intricate relationships among multiple targets make the conventional teacher-student network consistency loss unsuitable for multi-target tasks, leading to inaccurate feature capture and degraded performance. To address these issues, we propose a registration-based pseudo-label generation strategy that produces pseudo-labels more closely aligned with actual labels. Additionally, we introduce multiple guidance mechanisms into the teacher-student network to ensure correct optimization. Experimental results show that our approach significantly outperforms leading semi-supervised segmentation methods. Specifically, at a 50% label rate, our method improves the Dice coefficient by 6.9% over existing methods. © 2025 IEEE.

关键词： Multi-Object Registration Semi-Supervised Medical Image Segmentation

来源：评论

学校读者我要写书评

暂无评论

基于多域雷达回波数据融合的海面小目标分类网络模型

引用

电子与信息学报 2025年第3期47卷 696-706页

作者：赵子健许述文水鹏朗西安电子科技大学雷达信号处理全国重点实验室西安710071

海面小目标识别是海事雷达监视任务中一个重要且具有挑战性的问题。由于海面小目标类型多样、环境复杂多变,对其进行有效分类存在较大困难。在高分辨体制雷达下,海面小目标通常只占据一或几个距离单元,缺乏足够的空间散射结构信息,因此... 详细信息

海面小目标识别是海事雷达监视任务中一个重要且具有挑战性的问题。由于海面小目标类型多样、环境复杂多变,对其进行有效分类存在较大困难。在高分辨体制雷达下,海面小目标通常只占据一或几个距离单元,缺乏足够的空间散射结构信息,因此目标的雷达截面积(RCS)起伏和径向速度变化成为分类的主要依据。为此,该文提出一种基于多域雷达回波数据融合的分类网络模型,用于海面小目标的分类任务。由于不同域的数据具有其特殊的物理意义,因此该文构建了时域LeNet(T-LeNet)神经网络模块和时频特征提取神经网络模块,分别从雷达海面回波信号的幅度序列和时频分布(TFD)即时频图中提取特征。其中幅度序列主要反映了目标RCS的起伏特性,而时频图不仅反映RCS起伏特性,还能体现目标径向速度的变化。最后,利用IPIX,CSIR数据库和自测的无人机数据集构建了包括4种海面小目标的数据集:锚定漂浮小球、漂浮船只、低空无人机(UAV)和移动的快艇。实验结果表明所提方法具有良好的识别能力。

关键词：海面小目标目标分类多特征融合神经网络

来源：评论

学校读者我要写书评

暂无评论

JointSwinUNETR: An Efficient Feature-enhanced Architecture for small Intestine Cine MRI Segmentation

JointSwinUNETR: An Efficient Feature-enhanced Architecture f...

引用

2025 IEEE International conference on Acoustics, Speech, and signal processing, ICASSP 2025

作者： Wang, Yue Li, Wenhui Wang, Ziming Du, Taoli Ma, Ming Zhang, Mengchao College of Computer Science and Technology Jilin University Jilin Changchun China Department of Radiology China-Japan Union Hospital Jilin University Jilin Changchun China

ISBN: (纸本)9798350368741

The Cine MRI of the small intestine is a dynamic magnetic resonance imaging technique used to observe and evaluate small intestine motility. It captures sequential images of the organ in motion over time through rapid imaging. The Transformer architecture is highly effective at capturing long-range dependencies, enabling it to extract more comprehensive temporal information. As the image resolution provided by medical devices continues to improve, images now contain far more intricate details. However, most existing Transformer-based 3D segmentation models, constrained by computational resources, cannot directly handle high-resolution sequential Cine MRI data. In this paper, we propose an efficient joint training architecture for high-resolution 3D small intestine segmentation, consisting of a Local Sequential Encoder (LSE), a Global Multi-Scale Encoder (GMSE), and a Spatial Sequential Hybrid Decoder (SSHD). The LSE partitions high-resolution temporal images into multiple patches and uses a parameter-sharing mechanism to extract local detail features with a low computational cost. To prevent the loss of global features due to patch partitioning, the GMSE complements the LSE through joint training. Both encoders are trained simultaneously, the LSE captures local details, while the GMSE focuses on global, multi-scale features. Their combined outputs enable the model to balance fine-grained and large-scale information, improving segmentation accuracy. Furthermore, to address the periodic nature of small intestine peristalsis, we propose a dynamic consistency loss, which effectively quantifies motion amplitude differences between predicted and ground truth sequences. Experimental results on our dataset demonstrate that the proposed architecture significantly improves prediction accuracy compared to existing advanced 3D segmentation methods, while also delivering superior performance with low computational cost than recent advanced methods. © 2025 IEEE.

关键词： 3D segmentation cine MRI joint training small intestine

来源：评论

学校读者我要写书评

暂无评论

BiMA: Bidimensional multi-level attention embedded network for single-frame infrared small target detection

BiMA: Bidimensional multi-level attention embedded network f...

引用

International conference on Acoustics, Speech, and signal processing (ICASSP)

作者： He Deng Xiaojie Yin Xianmin Lan School of Computer Science and Technology Wuhan University of Science and Technology Wuhan China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

As the depth of the detection network increases, features of small targets become less pronounced, and the model may inadvertently favor background clutter, thereby reducing the efficiency of target detection. Hence, a bidimensional multilevel attention-embedded network, named BiMA, is designed to simultaneously tackle original images rich in detail and filtered images that remove superfluous non-target elements, serving as its dual-input framework. Subsequently, a multi-scale fusion is raised to harness multi-scale properties of these images, offering a thorough grasp of both high-level and low-level features. In addition, a multi-level attention is integrated to further improve the detection capability. Extensive qualitative and quantitative experiments prove that BiMA not only surpasses contemporary approaches in robustly detecting small targets across a range of challenging scenarios but also has superior performance, e.g., higher probabilities of detection, reduced false alarm rates, and broader areas under receiver operating characteristic curves.

关键词： Target tracking Target recognition Object detection Receivers signal processing Feature extraction Acoustics Clutter Speech processing

来源：评论

学校读者我要写书评

暂无评论

GREST: Ghost targets Removal Algorithm Using Multipath Angle Estimation

GREST: Ghost Targets Removal Algorithm Using Multipath Angle...

引用

International conference on Acoustics, Speech, and signal processing (ICASSP)

作者： Ryuhei Takahashi Pu Wang Information Technology R&D Center Mitsubishi Electric Corporation Kamakura Japan Mitsubishi Electric Research Laboratories (MERL) Cambridge MA USA

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

In this paper, we propose an algorithm for the removal of ghost targets based on angle estimation termed GREST (Ghost targets Removal using ESTAR). In the proposed GREST, the tentative targets reports, including the azimuth angles, is obtained through the conventional 3-D coherent integrations and signal detection at first. Then two-way angles, defined as a combination of direction-of-arrival (DOA) and direction-of-departure (DOD), are estimated using the ESTAR (Estimation of Two-way Angle by MIMO Radar), which was proposed by the authors recently. Subsequently, the ghost targets are removed based on the estimated angles by comparing DOD and DOA to finalize the target reports. The ESTAR method represents a novel approach to angle estimation that yields unambiguous results, even in the case of MIMO radars with transmit sparse arrays, which are commonly utilized in the standard millimeter-wave radars. Based on these angles, it is possible to distinguish between direct and multipath propagation, thereby enabling the removal of ghost targets. In this paper, we will present the detail of the proposed GREST and demonstrate its effectiveness in removing ghost targets using experimental data acquired in a situation where multipath propagation is occurring.

关键词： Direction-of-arrival estimation Three-dimensional displays MIMO radar Estimation signal processing algorithms US Department of Defense Speech processing Standards signal detection signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

Challenges in Infrared small-Target Detection: A Benchmark of YOLO Models on UAV and Bird Infrared Imagery

Challenges in Infrared Small-Target Detection: A Benchmark o...

引用

International conference on signal processing and Communication (ICSC)

作者： Swapnil Shahapurkar Pratiksha Angadi Chandni Kumari Prashant Uppar Prema T. Akkasaligar Department of Computer Science and Engineering KLE Tech University's Dr. MSSCET Belagavi India

ISBN: (数字)9798331511890

ISBN: (纸本)9798331511906

Detecting small objects in infrared (IR) images, such as birds and Unmanned Aerial Vehicles (UAVs), presents significant challenges due to their reduced size, limited pixel representation, and lack of distinct features. These challenges are further complicated in real-world environments where complex backgrounds, such as skies, clouds, and vegetation, combined with low signal-to-noise ratio (SNR) and low signal-to-clutter ratio (SCR), obscure the targets. Existing detection models often struggle to accurately detect small targets due to low contrast and indistinct edges. This highlights the need for detection techniques tailored to small object detection in infrared imagery. In this study, we evaluate the performance of various You Only Look Once (YOLO) models, focusing on their applicability and limitations for detecting small infrared objects. The novelty of this work lies in systematically analyzing the impact of preprocessing techniques and fine-tuning on detection performance. Fine-tuned YOLO models are evaluated for small target detection, with their strengths and weaknesses identified under diverse conditions. Insights are provided for adapting these models to enhance real-time surveillance systems in critical applications.

关键词： YOLO Training Adaptation models Shape Vegetation mapping Streaming media Birds Autonomous aerial vehicles Real-time systems signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

Design and Simulation of an FMCW/CW X-Band Radar for Detection/Classification of LSS targets

Design and Simulation of an FMCW/CW X-Band Radar for Detecti...

引用

International conference on Wireless Communications, signal processing and Networking (WiSPNET)

作者： A Amalorpava Selvi S P K. Babu Department of Electronics and Communication Engineering Periyar Maniammai Institute of Science and Technology Thanjavur India

ISBN: (数字)9798331524227

ISBN: (纸本)9798331524234

Low, small, and Slow (LSS) targets like Uncrewed Aerial Vehicles (UAVs) are becoming common due to their easy availability. They have various applications and sometimes when used for illegal operations pose a threat. The main types of LSS targets are bionic bird, quadcopter and helicopter. Monitoring such targets can be done by radars which can operate at night or day, foggy conditions, rainy/snowy weather. Due to their small size, low velocity and low altitude operation they are difficult to detect and identify. Radar Cross Section (RCS) of birds and some UAVs are in similar and it is therefore of little use. Micro-Doppler (m-D) signatures produced by micro motion by moving parts of the target will be of use as it is different for birds and drones. For birds it is flipping of the wings and for drones it will be from propellers. There are different types of radars where Frequency Modulated Continuous Wave (FMCW) radar can be used for detecting range and velocity and Continuous Wave (CW) radar can used for m-D images. In this research the above-mentioned radars at X-band are designed and simulated. It is found that the FMCW radar is accurately able to detect the distance of upto 5000 m and velocity of 25 m/s whereas the CW radar can capture the mD images.

关键词： Wireless communication Surveys Radar cross-sections Frequency modulation Radar detection Radar Radar imaging Birds Radar signal processing Drones

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：