检索结果-内蒙古大学图书馆

Propagating prior information with transformer for robust visual object tracking

MULTIMEDIA SYSTEMS 2024年第5期30卷 1-14页

作者： Wu, Yue Cai, Chengtao Yeo, Chai Kiat Harbin Engn Univ Sch Intelligent Sci & Engn Harbin 150001 Peoples R China Harbin Engn Univ Key Lab Intelligent Technol & Applicat Marine Equi Minist Educ Harbin 150001 Peoples R China Nanyang Technol Univ Sch Comp Sci & Engn Singapore 639798 Singapore

In recent years, the domain of visual object tracking has witnessed considerable advancements with the advent of deep learning methodologies. Siamese-based trackers have been pivotal, establishing a new architecture with a weight-shared backbone. With the inclusion of the transformer, attention mechanism has been exploited to enhance the feature discriminability across successive frames. However, the limited adaptability of many existing trackers to the different tracking scenarios has led to inaccurate target localization. To effectively solve this issue, in this paper, we have integrated a siamese network with the transformer, where the former utilizes ResNet50 as the backbone network to extract the target features, while the latter consists of an encoder and a decoder, where the encoder can effectively utilize global contextual information to obtain the discriminative features. Simultaneously, we employ the decoder to propagate prior information related to the target, which enables the tracker to successfully locate the target in a variety of environments, enhancing the stability and robustness of the tracker. Extensive experiments on four major public datasets, OTB100, UAV123, GOT10k and LaSOText demonstrate the effectiveness of the proposed method. Its performance surpasses many state-of-the-art trackers. Additionally, the proposed tracker can achieve a tracking speed of 60 fps, meeting the requirements for real-time tracking.

关键词： visual object tracking Siamese network Transformer Prior information

来源：评论

学校读者我要写书评

暂无评论

A Universal Event-Based Plug-In Module for visual object tracking in Degraded Conditions

引用

INTERNATIONAL JOURNAL OF COMPUTER VISION 2024年第5期132卷 1857-1879页

作者： Zhang, Jiqing Dong, Bo Fu, Yingkai Wang, Yuanchen Wei, Xiaopeng Yin, Baocai Yang, Xin Dalian Univ Technol Dalian Peoples R China Dalian Maritime Univ Dalian Peoples R China Princeton Univ Princeton NJ USA

Most existing trackers based on RGB/grayscale frames may collapse due to the unreliability of conventional sensors in some challenging scenarios (e.g., motion blur and high dynamic range). Event-based cameras as bioinspired sensors encode brightness changes with high temporal resolution and high dynamic range, thereby providing considerable potential for tracking under degraded conditions. Nevertheless, events lack the fine-grained texture cues provided by RGB/grayscale frames. This complementarity encourages us to fuse visual cues from the frame and event domains for robust object tracking under various challenging conditions. In this paper, we propose a novel event feature extractor to capture spatiotemporal features with motion cues from event-based data by boosting interactions and distinguishing alterations between states at different moments. Furthermore, we develop an effective feature integrator to adaptively fuse the strengths of both domains by balancing their contributions. Our proposed module as the plug-in can be easily applied to off-the-shelf frame-based trackers. We extensively validate the effectiveness of eight trackers extended by our approach on three datasets: EED, VisEvent, and our collected frame-event-based dataset FE141. Experimental results also show that event-based data is a powerful cue for tracking.

关键词： Event-based camera visual object tracking Multimodal fusion Plug-in module

来源：评论

学校读者我要写书评

暂无评论

SiamDMU: Siamese Dual Mask Update Network for visual object tracking

引用

IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE 2024年第2期8卷 1656-1669页

作者： Liu, Jing Wang, Han Ma, Chao Su, Yuting Yang, Xiaokang Tianjin Univ Sch Elect & Informat Engn Tianjin 300072 Peoples R China Shanghai Jiao Tong Univ AI Inst Shanghai 200240 Peoples R China

The Siamese trackers have demonstrated powerful performance in object tracking, which convert tracking into template matching. However, the trackers may fail to track the target whose appearance changes drastically. The intuitive solution of the problem is to accurately update the template in time. After analyzing the previous works, we find that a good template update mechanism is desired to generate a comprehensive and robust representation, to grasp and exploit the target appearance change trend, and to be easy to train. Therefore, we propose a novel tracker named Siamese Dual Mask Update (SiamDMU), which utilizes long-term motion and semantic information to generate the enhanced tracking results for template update. SiamDMU exploits the long-term motion information of target at certain time intervals to capture the trend of target appearance change. The core structure of template update in SiamDMU consists of two convolution layers and be easy to train with 19 videos. Experiments on VOT2016, VOT2018, OTB100, UAV123 and LaSOT, show that SiamDMU has competitive performance with state-of-the-art trackers with real-time speed.

关键词： Target tracking Feature extraction Semantics object tracking Videos visualization Frequency modulation visual object tracking Siamese tracker template update

来源：评论

学校读者我要写书评

暂无评论

Icg: intensity and color gradient operator on RGB images for visual object tracking

引用

visual COMPUTER 2024年第8期40卷 5797-5813页

作者： Dasari, Mohana Murali Gorthi, Rama Krishna Indian Inst Technol Dept Elect Engn Tirupati 517619 Andhra Pradesh India

The design of digital filters is now mostly automated with convolutional neural networks (CNNs). State-of-the-art works in tracking methods, including the well-known correlation and deep Siamese trackers, use features from such CNNs. However, deep learning requires huge data, high computational resources, and more training time. Hence, smart and simple alternative feature extraction strategies are needed in embedded applications. In this direction, a method is proposed for obtaining enriched "intensity and color gradient features" using the "three-dimensional gradient operator" on color images. This work considered the popular first-order gradient operator ([-1 0 1]) and outer product operator to generate various intensity and color gradients. The generated features contain rich information, including edges, color, and mid-level segmentation-like features. This simple yet effective operator does not involve any learning parameters. Despite not having learnable parameters, the proposed method's performance is comparable to lightweight learned CNNs such as MobileNet. The efficacy of the resultant features is demonstrated for the visual tracking task using well-known datasets in tracking, namely GOT-10k, LaSOT, OTB2015, UAV123, and VOT2018. The proposed features combined with other deep learning features have boosted the performance over baseline efficient convolution operator (ECO) tracker. This work will open new frontiers in designing hybrid features for visual tracking and other visual computation tasks. The code is available "https://***/dasari4321/ICG_***"

关键词： visual object tracking Gradient image features Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Learning Attention Through Hierarchical Architecture for visual object tracking

引用

IEEE SIGNAL PROCESSING LETTERS 2024年 31卷 186-190页

作者： Wang, Qinghui Yang, Peng Dou, Lei Nanjing Univ Sci & Technol Natl Key Lab Transient Phys Nanjing 210094 Peoples R China

Various attention mechanisms have been extensively adopted in visual object tracking networks. However, the research of Siamese networks with convolutional attention mechanisms as the main architecture remains limited. In this letter, we propose a novel Siamese tracker that leverages convolutional attention and a hierarchical architecture. By utilizing the deformable convolution, we expand the receptive field to capture more contextual information. In addition, our hierarchical architecture optimizes the attention by hierarchically adjusting the focus of attention based on the significance of features at different levels. On the other hand, we decompose the step of feature compression during the feature-to-result transformation process and integrate it in segments with the feature fusion process to address the information loss and underfitting issues. Extensive experimental results demonstrate that our tracker achieves state-of-the-art performance on multiple benchmarks including LaSOT, trackingNet, UAV123, and NFS while maintaining a high processing speed of 40 FPS.

关键词： Attention mechanism deep learning siamese tracker visual object tracking

来源：评论

学校读者我要写书评

暂无评论

Learning Feature Restoration Transformer for Robust Dehazing visual object tracking

引用

INTERNATIONAL JOURNAL OF COMPUTER VISION 2024年第12期132卷 6021-6038页

作者： Xu, Tianyang Pan, Yifan Feng, Zhenhua Zhu, Xuefeng Cheng, Chunyang Wu, Xiao-Jun Kittler, Josef Jiangnan Univ Sch Artificial Intelligence & Comp Sci Wuxi 214122 Peoples R China Univ Surrey Sch Comp Sci & Elect Engn Guildford GU2 7XH England Univ Surrey Ctr Vis Speech & Signal Proc Guildford GU27XH England

In recent years, deep-learning-based visual object tracking has obtained promising results. However, a drastic performance drop is observed when transferring a pre-trained model to changing weather conditions, such as hazy imaging scenarios, where the data distribution differs from that of a natural training set. This problem challenges the open-world practical applications of accurate target tracking. In principle, visual tracking performance relies on the discriminative degree of features between the target and its surroundings, rather than the image-level visual quality. To this end, we design a feature restoration transformer that adaptively enhances the representation capability of the extracted visual features for robust tracking in both natural and hazy scenarios. Specifically, a feature restoration transformer is constructed with dedicated self-attention hierarchies for the refinement of potentially contaminated deep feature maps. We endow the feature extraction process with a refinement mechanism typically for hazy imaging scenarios, establishing a tracking system that is robust against foggy videos. In essence, the feature restoration transformer is jointly trained with a Siamese tracking transformer. Intuitively, the supervision for learning discriminative and salient features is facilitated by the entire restoration tracking system. The experimental results obtained on hazy imaging scenarios demonstrate the merits and superiority of the proposed restoration tracking system, with complementary restoration power to image-level dehazing. In addition, consistent advantages of our design can be observed when generalised to different video attributes, demonstrating its capacity to deal with open-world scenarios.

关键词： visual object tracking Dehazing system Siamese tracker Feature restoration

来源：评论

学校读者我要写书评

暂无评论

Regression-Selective Feature-Adaptive Tracker for visual object tracking

引用

IEEE TRANSACTIONS ON MULTIMEDIA 2023年 25卷 5444-5457页

作者： Zhou, Ze Sun, Quansen Li, Hongjun Li, Chaobo Ren, Zhenwen Nanjing Univ Sci & Technol Sch Comp Sci & Engn Nanjing 210094 Jiangsu Peoples R China Nantong Univ Sch Informat Sci & Technol Nantong 226019 Jiangsu Peoples R China Southwest Univ Sci & Technol Sch Natl Def Sci & Technol Mianyang 621010 Peoples R China

As a challenging visual task, visual object tracking has recently been composed of the classification and regression subtasks. The anchor-free regression network gets rid of the dependence on the anchors, but the redundant range makes it usually regress some samples involving non-target information. Evenly dividing a target by the regular receptive field often causes ambiguous target localization. To address these issues, we propose a regression-selective feature-adaptive tracker (RSFA), where the regression-selective subnetwork can not only free the regression task from anchors, but can also select more effective regression samples using the refined criterion. The proposed feature-adaptive strategy concentrates the classification subnetwork on target feature extraction via adaptively modifying the receptive field, and the attached centrality branch offers a correction for target localization by exploiting the spatial information. Additionally, the designed online update mechanism realizes the tracker's online optimization, improving robustness against target deformation. Extensive experiments are conducted on challenging benchmarks, including GOT10 K, OTB2015, UAV123, NFS, VOT2018, VOT2019 and VOT2020-ST. Our tracker achieves satisfactory tracking results, and the evaluations of its tracking performance rank first or second in comparison with the state-of-the-art tracking algorithms.

关键词： visual object tracking Refined criterion Adaptive receptive field Centrality Online update

来源：评论

学校读者我要写书评

暂无评论

An Energy-Efficient visual object tracking Processor Exploiting Domain-Specific Features

引用

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS II-EXPRESS BRIEFS 2024年第5期71卷 2794-2798页

作者： Gong, Yuchuan Guo, Hongtao Liu, Xiyuan Zheng, Jingxiao Zhang, Teng Que, Luying Jia, Conghan Ou, Guangbin Jiao, Xiben Liu, Zherong Chang, Liang Zhou, Liang Zhou, Jun Univ Elect Sci & Technol China Sch Informat & Commun Engn Chengdu 611731 Sichuan Peoples R China Univ Elect Sci & Technol China Chengdu 611731 Peoples R China

Nowadays, visual object tracking (VOT) has become a key technology in many AI applications such as intelligent surveillance and mobile robots. In the past, general AI accelerators have been developed and used for accelerating VOT. However, as the domain-specific knowledge is not well utilized, the energy efficiency of general AI accelerators is limited when accelerating VOT. To address this issue, in this brief, an energy-efficient VOT processor is proposed by exploiting diverse domain-specific features of VOT processing. Several techniques have been proposed to improve the energy efficiency and processing speed, including an efficient sparse computing architecture with channel-wise data and weight compression, a Siamese-core processing technique and a bounding box (bbox) early-drop technique. Implemented and fabricated with a 55nm CMOS technology, the proposed VOT processor achieves high energy efficiency (15.08 TOPS/W), high frame rate (69.8 fps) with high accuracy, outperforming several state-of-the-art designs.

关键词： Computational efficiency Computer architecture Clocks Kernel Energy efficiency Task analysis Artificial neural networks visual object tracking energy efficiency hardware accelerator

来源：评论

学校读者我要写书评

暂无评论

Efficient Siamese model for visual object tracking with attention-based fusion modules

引用

SIGNAL IMAGE AND VIDEO PROCESSING 2024年第11期18卷 7801-7810页

作者： Zhou, Wenjun Liu, Yao Wang, Nan Liang, Dong Peng, Bo Southwest Petr Univ Sch Comp Sci & Software Engn Chengdu Peoples R China Nanjing Univ Aeronaut & Astronaut Nanjing Peoples R China

In the realm of visual object tracking, achieving precise target localization poses a significant challenge. While Siamese trackers based on shallow networks often prioritize speed, they tend to suffer from subpar localization performance. This can be attributed to the inadequacy of features extracted by shallow networks, making it challenging to accurately pinpoint the target within intricate backgrounds. This paper presents an efficient Siamese model architecture for visual object tracking, aimed at mitigating the challenge of suboptimal feature extraction often observed in shallow network designs. On the basis of the Siamese network, we devise the Channel Attention Feature Enhancement Fusion Module and Spatial-Channel Attention Feature Enhancement Fusion Module to bolster feature extraction while suppressing background noise. Furthermore, we enhance the model's resilience to disturbances by combining Siamese classification scores with those from an online classifier and to maintain high-quality image features during tracking, we introduce an online updating strategy for refreshing the template *** results across four publicly available datasets consistently demonstrate the outstanding performance of the proposed Siamese model *** code is available at https://***/lejunXTS/ESM-VOTAFM/tree/master.

关键词： visual object tracking Siamese network Attention mechanism Template updating Online classifier

来源：评论

学校读者我要写书评

暂无评论

visual object tracking and Servoing Control of a Nano-Scale Quadrotor:System,Algorithms,and Experiments

引用

IEEE/CAA Journal of Automatica Sinica 2021年第2期8卷 344-360页

作者： Yuzhen Liu Ziyang Meng Yao Zou Ming Cao the Department of Precision Instrument Tsinghua UniversityBeijing 100084China the School of Automation and Electrical Engineering University of Science and Technology BeijingBeijing 10083China the Institute of Engineering and Technology University of GroningenGroningen 9747AGThe Netherlands

There are two main trends in the development of unmanned aerial vehicle(UAV)technologies:miniaturization and intellectualization,in which realizing object tracking capabilities for a nano-scale UAV is one of the most challenging *** this paper,we present a visual object tracking and servoing control system utilizing a tailor-made 38 g nano-scale quadrotor.A lightweight visual module is integrated to enable object tracking capabilities,and a micro positioning deck is mounted to provide accurate pose *** order to be robust against object appearance variations,a novel object tracking algorithm,denoted by RMCTer,is proposed,which integrates a powerful short-term tracking module and an efficient long-term processing *** particular,the long-term processing module can provide additional object information and modify the short-term tracking model in a timely ***,a positionbased visual servoing control method is proposed for the quadrotor,where an adaptive tracking controller is designed by leveraging backstepping and adaptive *** and accurate object tracking is achieved even under *** results are presented to demonstrate the high accuracy and stability of the whole tracking system.

关键词： Nano-scale quadrotor nonlinear control positionbased visual servoing visual object tracking

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：