检索结果-内蒙古大学图书馆

Automatic Weight Allocation: optimizing remote sensing image retrieval from contrastive learning perspective

Multimedia Systems 2025年第3期31卷 1-19页

作者： Wang, Sijia Ge, Yun Liu, Qiyang Zeng, Yan School of Software Nanchang Hangkong University Nanchang China Jiangxi Province Key Laboratory of Image Processing and Pattern Recognition Nanchang China

Traditional supervised learning methods achieve remarkable performance in high-resolution remote sensing image retrieval, but are limited by the dependence on large-scale annotated images. Contrastive learning can leverage unlabeled images to learn powerful visual features, demonstrating its potential in many unsupervised tasks. Moreover, hash algorithms show significant potential in the field of image retrieval with their advantages in efficiency and storage. Therefore, we propose the Contrastive Hashing Framework based on Automatic Weight Allocation. The framework employs a two-stage training strategy. In the feature learning stage, we propose the Automatic Weighted Contrastive Loss (AWCLoss). It incorporates Gaussian weighting and dynamic adjustment strategies to improve loss functions, enabling them to focus on the distinctiveness and importance of samples. Gaussian weighting assigns different weight values based on the similarity of sample pairs, enhancing the learning of critical sample pairs. Meanwhile, the dynamic adjustment strategy sets a threshold to identify hard negative samples and then adjusts the weight values to weaken the model from being disturbed by hard negative samples. In the hashing learning stage, a hashing layer is added to the end of the network, which converts high-dimensional representations into hash codes. The Quantization loss is introduced to learn the hash codes so that the semantic similarity structure between data can be preserved in hamming space. Additionally, the AWCLoss is utilized to enhance the discriminative power of the hash codes. Extensive experiments on three remotely sensed datasets UCM, AID and NWPU-RESISC45 have demonstrated the significant superiority of our approach in remote sensing image retrieval. Our source code is available at https://***/WANGSJ77/AWCH .

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Research Mode Based Evolutionary Algorithm for Many-Objective Optimization

引用

Chinese Journal of Electronics 2019年第4期28卷 764-772页

作者： CHEN Guoyu LI Junhua Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Nanchang Hangkong University

The development of algorithms to solve Many-objective optimization problems(MaOPs) has attracted significant research interest in recent *** various types of Pareto front(PF) is a daunting challenge for evolutionary algorithm. A Research mode based evolutionary algorithm(RMEA) is proposed for many-objective optimization. The archive in the RMEA is used to store non-dominated solutions that can reflect the shape of the PF to guide the reference vector *** concerning the population is collected, once the number of non-dominated solutions reaches its limit after many generations without exceeding a given threshold, RMEA introduces a research mode that generates more reference vectors to search through the solutions. The proposed algorithm showed competitive performance with four state-of-the-art evolutionary algorithms in a large number of experiments.

关键词： Many-objective optimization Reference vector Research mode Evolutionary algorithm

来源：评论

学校读者我要写书评

暂无评论

Segmentation of the Ventricle Membranes in Short-Axis Sequences by Optical Flow Base on DLSRE Model

引用

Chinese Journal of Electronics 2021年第3期30卷 460-470页

作者： LI Lin WU Hengfei LI Junhua Department of Electronics and Information Engineering Bozhou University Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Nanchang Hangkong University

Recent studies have pointed out that the boundary of the extracted ventricle membranes is unsmooth, and the segmentation of the cardiac papillary muscle and trabecular muscle do inconformity the clinical requirements. To address these issues, this paper proposes an automatic segment algorithm for continuously extracting ventricle membranes boundary, which adopts optical flow field information and sequential images information. The images are cropped by frame difference method, which according to the continuity of adjacent slices of cardiac MRI images. The roughly boundary of epicardium is extracted by the Double level set region evolution(DLSRE) model, which combines image global information, local information and edge information. The ventricle endocardium and epicardial contours are tracked according to the optical flow field information between image sequences. The segmentation results are optimized by Delaunay triangulation algorithm. The experimental results demonstrate that the proposed method can improve the accuracy of segmenting the ventricle endocardium and epicardium contours, and segment the contour of the smooth ventricle membrane edge that meets the clinical definition.

关键词： edge information image sequences automatic segment algorithm sequential images information clinical requirements medical image processing cardiology optical flow base extracted ventricle membranes roughly boundary local information ventricle endocardium trabecular muscle optical flow field information continuously extracting ventricle membranes boundary DLSRE model biomedical MRI smooth ventricle membrane edge image global information mesh generation muscle short-axis sequences cardiac MRI images image segmentation cardiac papillary muscle frame difference method region evolution model

来源：评论

学校读者我要写书评

暂无评论

Deep Learning in Palmprint recognition-A Comprehensive Survey

arXiv

引用

arXiv 2025年

作者： Gao, Chengrui Yang, Ziyuan Jia, Wei Leng, Lu Zhang, Bob Teoh, Andrew Beng Jin College of Computer Science Sichuan University Chengdu610065 China Singapore School of Computer and Information Hefei University of Technology Hefei China Jiangxi Provincial Key Laboratory of Image Processing and Pattern Recognition Nanchang Hangkong University Nanchang China Pattern Analysis and Machine Intelligence Group Department of Computer and Information Science University of Macau Taipa China School of Electrical and Electronic Engineering College of Engineering Yonsei University Seoul Korea Republic of

Palmprint recognition has emerged as a prominent biometric technology, widely applied in diverse scenarios. Traditional handcrafted methods for palmprint recognition often fall short in representation capability, as they heavily depend on researchers' prior knowledge. Deep learning (DL) has been introduced to address this limitation, leveraging its remarkable successes across various domains. While existing surveys focus narrowly on specific tasks within palmprint recognition-often grounded in traditional methodologies-there remains a significant gap in comprehensive research exploring DL-based approaches across all facets of palmprint recognition. This paper bridges that gap by thoroughly reviewing recent advancements in DL-powered palmprint recognition. The paper systematically examines progress across key tasks, including region-of-interest segmentation, feature extraction, and security/privacy-oriented challenges. Beyond highlighting these advancements, the paper identifies current challenges and uncovers promising opportunities for future research. By consolidating state-of-the-art progress, this review serves as a valuable resource for researchers, enabling them to stay abreast of cutting-edge technologies and drive innovation in palmprint recognition. Copyright © 2025, The Authors. All rights reserved.

关键词： Biometrics

来源：评论

学校读者我要写书评

暂无评论

GFENet: Enabling Attention and Adaptive Feature Fusion in Real-time Object Detection 2

GFENet: Enabling Attention and Adaptive Feature Fusion in Re...

引用

2nd International Symposium on Electrical, Electronics and Information Engineering, ISEEIE 2022

作者： Pan, Chaolin Zhu, Xiaoyang Chu, Jun Nanchang Hangkong University Department of Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Nanchang China

ISBN: (数字)9781665468749

ISBN: (纸本)9781665468749

Real-time object detection has been a critical component of onboard instrumentation for next-generation autonomous driving. To enable safety and reliability in autonomous driving systems, we must continue advancing its real-time object detection accuracy for future autonomous driving systems in complex environments. Existing works in real-time object detection cannot accurately detect objects with similar appearances and in complex environments. In particular, existing approaches missed many essential features because of the loss of multi-scale contextual features. Therefore, they cannot differentiate objects that have a similar appearance, a low-resolution, and a small size. To solve these issues, and to advance real-time object detection accuracy, we develop GFENet (Global Feature Enhancement Network), an end-to-end real-time object detection framework. First, we introduce a convolutional attention module that integrates spatial and channel-wise features and eliminates ineffective features to enable a strong feature representation of the network, followed by a lightweight residual module that further advances the feature representation by highlighting salient changes in features. Second, we build an adaptive feature fusion module to adaptively learn weights of multi-scale contextual features across layers, aiming to strengthen weights for small-size objects. We comprehensively evaluate GFENet on two leading object detection datasets-PASCAL VOC and MS COCO. The evaluation results (mAP of 82.9 and FPS of 31 on PASCAL VOC;mAP of 39.0 and FPS of 23 on MS COCO) show the outperformance of GFENet compared to competing baseline methods. © 2022 IEEE.

关键词： Instruments Object detection Detectors Feature extraction Real-time systems Safety Reliability

来源：评论

学校读者我要写书评

暂无评论

Object Detector based on Enhanced Multi-scale Feature Fusion Pyramid Network 5

Object Detector based on Enhanced Multi-scale Feature Fusion...

引用

5th IEEE Advanced Information Technology, Electronic and Automation Control Conference, IAEAC 2021

作者： Zhao, Luan Zhang, Xiaofeng Nanchang Hangkong University Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition School of Information Engineering Nanchang China

Constructing the pyramidal architecture for the feature is currently a very effective way to obtain feature information of objects at different scales. Although the feature pyramid can realize the recognition and detection of multi-scale objects in the object detection task well, it still has some limitations. Since the feature information of different levels is often not from the same layer of the network, it is difficult to obtain the feature of different objects information at a certain scale from a certain level feature map of the pyramid network. To solve this problem, we present a novel object detection architecture, named Enhanced Multi-scale Feature Fusion Pyramid Network (EMFFPNet). Our network consists of Enhanced Multi-scale Feature Fusion Module (EMFFM) and Predictor Optimization Module (POM). In EMFFM, Features at different levels can be fused into the Enhanced features as outputs, which are more representative and deterministic. In order to enable the enhanced features to play their respective roles in the pyramid network, we assign different weights to fusion features of different levels in POM. We perform the experiments on the COCO detection benchmark. The experimental results indicate that the performance of our model is much better than the state-of-the-art model. © 2021 IEEE.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Dense Connected Residual Generative Adversarial Network for Single image Deblurring 5

Dense Connected Residual Generative Adversarial Network for ...

引用

5th IEEE Advanced Information Technology, Electronic and Automation Control Conference, IAEAC 2021

作者： Qian, Pan Wu, Yeyun Zhang, Xiaofeng Nanchang Hangkong University School of Information Engineering Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Nanchang China

Recently, deep convolutional learning has been applied to image deblurring, which greatly improves the performance of single-image blind deblurring algorithms. However, most deep image deblurring models based on convolutional neural networks do not make full use of the information of low-level spatial convolutions, so they achieve relatively poor performance. In this paper, we improved the generator of the Generative Adversarial Network (GAN) and proposed dense networks and residual networks to solve this problem in image deblurring. The network makes full use of the hierarchical features of all convolutional layers. Specifically, we extract the local features of the convolutional layer through the proposed densely connected convolutional layer. Moreover, the dense connections of the residual dense network can connect all layers from the previous state to the current state. Then using local jump connection and feature fusion, it can adaptively learn more effective features from the previous and current local features, making the trained network more extensive. After fully obtaining the dense local features, the global residual connection is used to learn the global feature information in an overall manner. Experimental comparison and verification have proved the effectiveness of our method and achieved excellent results both quantitatively and qualitatively. © 2021 IEEE.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

An efficient multiparty quantum secret sharing scheme using a single qudit

引用

Chinese Physics B 2023年第8期32卷 161-170页

作者：胡文文熊邦书周日贵 School of Information Engineering Nanchang Hangkong UniversityNanchang 330063China Key Laboratory of Image Processing and Pattern Recognition of Jiangxi Province Nanchang Hangkong UniversityNanchang 330063China College of Information Engineering Shanghai Maritime UniversityShanghai 201306China

The aim of quantum secret sharing,as one of most promising components of quantum cryptograph,is one-tomultiparty secret communication based on the principles of quantum *** this paper,an efficient multiparty quantum secret sharing protocol in a high-dimensional quantum system using a single qudit is *** participant's shadow is encoded on a single qudit via a measuring basis encryption method,which avoids the waste of qudits caused by basis *** analysis indicates that the proposed protocol is immune to general attacks,such as the measure-resend attack,entangle-and-measure attack and Trojan horse *** to former protocols,the proposed protocol only needs to perform the single-qudit measurement operation,and can share the predetermined dits instead of random bits or dits.

关键词： quantum secret sharing high-dimensional measurement basis encrypted security analysis

来源：评论

学校读者我要写书评

暂无评论

An Improved Embedding Model for Zero-Shot Classification based on Attention Mechanism 5

An Improved Embedding Model for Zero-Shot Classification bas...

引用

5th IEEE Advanced Information Technology, Electronic and Automation Control Conference, IAEAC 2021

作者： Wu, Zongfeng Li, Qishen Yu, Xiao Peng, Xu Nanchang Hangkong University School of Information Engineering Nanchang Jiangxi China Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Nanchang Jiangxi China

It is well-known that the auxiliary information plays a key role in zero-shot classification. However, most of the existing popular methods do not make effective use of auxiliary information. To address this issue, we propose an improved embedding model for zero-shot classification based on attention mechanism, called EMAM. In the proposed EMAM, we first add an attention mechanism to effectively extract the key information of auxiliary information in zero-shot classification. Then optimizes the objective function to improve the recognition rate of this model. Finally the experimental comparison is implemented on the standard zero-shot learning datasets. The experimental results demonstrate that our proposed EMAM not only verifies its validity, but also achieves good results. © 2021 IEEE.

关键词： Zero-shot learning

来源：评论

学校读者我要写书评

暂无评论

Cross-GAN: Unsupervised image-to-image Translation 6

Cross-GAN: Unsupervised Image-to-Image Translation

引用

6th IEEE Information Technology and Mechatronics Engineering Conference, ITOEC 2022

作者： Peng, Xu Li, Qishen Wu, Taida Yuan, Sihao Nanchang Hangkong University School of Information Engineering Jiangxi Nanchang China Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Jiangxi Nanchang China

ISBN: (纸本)9781665431859

In recent years, unsupervised image-to-image translation is a topic in computer graphics. This paper proposes a new architecture to translate the image through the idea of sharing and decoupling, termed as Cross-GAN, which can better extract the style in the process of translation, and does not change the original outline and background. We improve the cycle consistency loss and reconstruction loss for the unique properties of the model to make the image more realistic. Because GAN[1] only generates the target style image by inputting the content image, and the information of the target domain image is not used in the training process. We input style images and content images through cross-input. The style images play a guiding role in the generator, and the effect of image are better by fusing style features and content features. Due to the fact that the image of the target domain is input, it is easy to produce the phenomenon of over-fitting. In order to prevent over-fitting training, we improve the cycle consistency loss and reuse the input encoder to make the model more compact. © 2022 IEEE.

关键词： image enhancement

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：