Article
Computer Science, Information Systems
Jingjing Xiao, Mourad Oussalah
Summary: Color-based particle filters are a popular method for tracking targets that undergo rapid appearance changes. Traditional updates without contextual learning can distort the model and lead to target loss. A new algorithm proposed in this paper uses environmental information to update the scale of the tracker and the reference appearance model for object tracking in video sequences. The algorithm differentiates foreground and background particles based on matching score and investigates a roaming phenomenon that affects estimation accuracy. Tests show the feasibility of the proposal and lay foundations for further research on complex visual tracking problems.
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
(2021)
Article
Engineering, Electrical & Electronic
Mario I. Chacon-Murguia, Andrea Rivero-Olivas, Juan A. Ramirez-Quintana
Summary: The method proposed in this paper involves using a fuzzy weighted color histogram neural network to estimate target position and update appearance model, combining shape and color information. Through online dynamic training and correction stages, the accuracy and stability of target tracking are improved.
SIGNAL IMAGE AND VIDEO PROCESSING
(2021)
Article
Automation & Control Systems
Hui Li, Yapeng Liu, Xiaoguo Liang, Yongfeng Yuan, Yuanzhi Cheng, Guanglei Zhang, Shinichi Tamura
Summary: This paper proposes a tracking-by-detection framework for multi-object tracking (MOT) that detects objects in each frame and identifies associations with objects in the previous frame. A deep association network is used to match object features and calculate associations to achieve accurate tracking. The framework addresses the problem of missing and partial detection and is particularly suitable for solving object ID switch caused by occlusion, entering and leaving of objects.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Geochemistry & Geophysics
Yidan Nie, Chunjiang Bian, Ligang Li
Summary: This study proposes a novel tracker for remote sensing tracking, which utilizes a multidimensional information-aware module and a temporal motion compensation mechanism. Experimental results show that our method outperforms other tracking models, especially in occlusion scenarios.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Weiming Hu, Qiang Wang, Li Zhang, Luca Bertinetto, Philip H. S. Torr
Summary: In this article, SiamMask, a framework for real-time visual object tracking and video object segmentation, using the same simple method, is introduced. The offline training procedure of popular fully-convolutional Siamese approaches is improved by adding a binary segmentation task. Once the offline training is completed, SiamMask can perform visual object tracking and segmentation at high frame-rates with only a single bounding box for initialization. The framework can also handle multiple object tracking and segmentation by re-using the multi-task model in a cascaded fashion.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Longfeng Shen, Xiaoxiao Wang, Lei Liu, Bin Hou, Yulei Jian, Jin Tang, Bin Luo
Summary: In this paper, a cooperative low-rank graph model algorithm is proposed to suppress background clutter by decomposing the input dual-modal features and using regularized graph learning. This improves the robustness of tracking performance.
Article
Computer Science, Artificial Intelligence
Huai Qin, Changqian Yu, Changxin Gao, Nong Sang
Summary: In this paper, a Detection to Tracking (D2T) framework is proposed to effectively transfer existing object detection methods to visual tracking task. The framework addresses the differences in object definition and temporal dimension through a general-to-specific network and a temporal strategy. The D2T framework is the first universal framework that directly applies deep learning based object detectors to visual tracking task.
PATTERN RECOGNITION
(2022)
Article
Multidisciplinary Sciences
Zhongyi Hu, Jingjing Shao, Feiyan Nie, Zhenzhen Luo, Changzu Chen, Lei Xiao
Summary: A robust online learning ship tracking algorithm based on the Siamese network and the region proposal network is proposed to address the inaccurate estimation of the target ship's motion state caused by frequent occlusions in inland river scenes. The algorithm combines off-line and online learning, and utilizes an occlusion determination mechanism and a global search mechanism to avoid tracking drift. Additionally, an efficient adaptive online update strategy is introduced to improve template degradation in the tracking process.
SCIENTIFIC REPORTS
(2023)
Article
Mathematics, Interdisciplinary Applications
Wen Zhong, Jian Xiong, Anping Lin, Lining Xing, Feilong Chen, Yingwu Chen
Summary: The study introduces a flexible ensemble framework ASES that enhances the performance of solving multi-objective optimization problems by embedding different MOEAs. By recording large-scale nondominated solutions in a big archive and developing an efficient strategy to update the archive, the efficiency of the algorithm is improved.
Article
Automation & Control Systems
Tao Ying, Huaicheng Yan, Zhichen Li, Kaibo Shi, Xiangsai Feng
Summary: Loop closure detection is a crucial part of SLAM, and using image histogram and key region covariance matrix matching method can improve the accuracy and recall rate of loop closure detection.
INTERNATIONAL JOURNAL OF CONTROL AUTOMATION AND SYSTEMS
(2021)
Article
Engineering, Electrical & Electronic
Xiang Xu, Jian Zhao, Jianmin Wu, Furao Shen
Summary: This paper proposes a new long-term video object tracking framework that simplifies the framework by carefully selecting advanced trackers. It combines two mainstream short-term tracking pipelines and introduces a global re-detector. Experimental results show remarkable results on seven long-term VOT datasets by utilizing the capabilities of existing methods instead of designing new neural networks. By introducing a continuous adjustable speed control parameter, the tracker achieves 20+FPS with only a small performance loss. The refine module improves bounding box estimations and outputs segmentation masks, enabling the framework to handle video object segmentation tasks using only VOT trackers. A trade-off between time and accuracy is achieved on two representative VOS datasets by using bounding boxes as the initial input.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Engineering, Civil
Xiaomeng Cao, Jian Lan, X. Rong Li, Yu Liu
Summary: This paper introduces a novel approach for automotive radar-based extended object tracking, which jointly estimates the kinematic state and extension of a vehicle, using a rectangular shape to describe the vehicle and partitioning the area to simplify the scattering center distribution modeling. The proposed method demonstrates its effectiveness through simulated and real data.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Chemistry, Analytical
Lesole Kalake, Yanqiu Dong, Wanggen Wan, Li Hou
Summary: This paper proposes a method that combines a convolutional neural network and a histogram of oriented gradient descriptor to improve the accuracy of multi-object tracking in video surveillance. Experimental results demonstrate that the proposed technique achieves better detection rate and data associations than the state-of-the-art approach.
Article
Environmental Sciences
Haotian Yang, Bin Zhou, Lixin Wang, Qi Wei, Feng Ji, Rong Zhang
Summary: This paper proposes a cascaded adaptive vector tracking method based on the KF+EKF architecture, implemented through the signal tracking module and the navigation solution module. The method adjusts the measurement and process noise covariance matrices based on the signal C/N-0 ratio and innovation sequence to improve positioning accuracy.
Article
Geography, Physical
Xinbo Qiao, Yongqiang Zhao, Lulu Chen, Seong G. Kong, Jonathan Cheung-Wai Chan
Summary: This paper introduces a feature descriptor Mosaic Gradient Histogram (MGH) for real-time object tracking in DoFP infrared polarization imaging, which does not require demosaicing and outperforms conventional feature descriptors in performance metrics.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Lei Tong, Zhihua Liu, Zheheng Jiang, Feixiang Zhou, Long Chen, Jialin Lyu, Xiangrong Zhang, Qianni Zhang, Abdul Sadka, Yinhai Wang, Ling Li, Huiyu Zhou
Summary: Depression is a common mental health disorder, and many sufferers do not seek help due to shame or lack of awareness. This paper proposes a novel classifier, CBPT, which can accurately detect depression by mining online social behaviors. The results show that the framework has promising potential for identifying Twitter users with depression.
IEEE TRANSACTIONS ON AFFECTIVE COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Zhihua Liu, Lei Tong, Long Chen, Zheheng Jiang, Feixiang Zhou, Qianni Zhang, Xiangrong Zhang, Yaochu Jin, Huiyu Zhou
Summary: Brain tumor segmentation is a challenging problem in medical image analysis, and deep learning methods have shown promising results in this field. This survey provides a comprehensive study of recently developed deep learning techniques for brain tumor segmentation, covering technical aspects such as network architecture design, segmentation under imbalanced conditions, and multi-modality processes. It also offers insightful discussions for future development directions.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Xiuxiu Ren, Xiangwei Zheng, Lizhen Cui, Gang Wang, Huiyu Zhou
Summary: This paper proposes a novel Asymmetric Similarity-Preserving Discrete Hashing (ASPDH) method for learning compact binary codes for image retrieval. The method achieves simultaneous similarity preservation and hash code learning, leading to improved discriminative capability of the learned binary codes.
APPLIED INTELLIGENCE
(2023)
Article
Geochemistry & Geophysics
Fang Chen, Heiko Balzter, Feixiang Zhou, Peng Ren, Huiyu Zhou
Summary: In this article, we propose an effective segmentation framework called DGNet for oil spill segmentation in SAR images. Our framework incorporates the intrinsic distribution of backscatter values in SAR images and utilizes two deep neural modules for inference and generation. Experimental evaluations show that DGNet achieves accurate oil spill segmentation.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Automation & Control Systems
Ashima Anand, Amit Kumar Singh, Huiyu Zhou
Summary: This article presents a visible and imperceptible medical data hiding technique, named ViMDH, which can prevent illegal copying and forgery of medical records. The technique involves marking the carrier image and implementing imperceptible data hiding to ensure the security of medical records.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Automation & Control Systems
Dajun Du, Changda Zhang, Xue Li, Minrui Fei, Huiyu Zhou
Summary: This article proposes a linear event-triggered dynamic watermarking scheme to enhance the cyberattack detection capability of networked control systems. It investigates the limitations of conventional dynamic watermarking schemes for event-triggered state estimation-based NCSs and designs a new scheme based on symmetric key encryption. The security property against generalized replay attacks is also discussed. Experimental results validate the effectiveness of the proposed scheme.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Xue Qin, Yi Niu, Huiyu Zhou, Xiaojie Li, Weikuan Jia, Yuanjie Zheng
Summary: Accurate identification of driver's drowsiness state through Electroencephalogram (EEG) signals can effectively reduce traffic accidents. This study proposes a fusion model based on tree Federated Learning (FL) and Convolutional Neural Network (CNN) to construct an efficient and accurate privacy-preserving drowsiness monitoring system. The method achieves higher accuracy, F1-score, and AUC than traditional methods, while better protecting privacy and improving communication efficiency.
INTERNATIONAL JOURNAL OF NEURAL SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Faming Xu, Chen Qiao, Huiyu Zhou, Vince D. Calhoun, Julia M. Stephen, Tony W. Wilson, Yuping Wang
Summary: Multi-paradigm deep learning models have the potential to analyze dynamic functional connectivity (dFC) by integrating complementary information. This paper proposes a multi-paradigm fusion-based explainable deep sparse autoencoder (MF-EDSAE) to address the limitations of existing models. The MF-EDSAE exhibits better performance than a single-paradigm deep sparse autoencoder (DSAE) in detecting significant differences in dFC during brain development.
Article
Multidisciplinary Sciences
Felipe Montes, Martha Blanco, Andres F. Useche, Sharon Sanchez-Franco, Carlos Caro, Lei Tong, Jie Li, Huiyu Zhou, Jennifer M. Murray, Olga L. Sarmiento, Frank Kee, Ruth F. Hunter
Summary: In this study, the researchers examined the influence of social networks on social norms related to adolescent smoking in specific school settings in Northern Ireland and Colombia. The results showed that students were more likely to be friends with others who had social norms against smoking. However, students with social norms favorable towards smoking had more friends with similar views than those with perceived norms against smoking, highlighting the importance of network thresholds.
SCIENTIFIC REPORTS
(2023)
Article
Computer Science, Information Systems
Zhuo Kuang, Zengqiang Yan, Huiyu Zhou, Li Yu
Summary: This paper presents a weakly-supervised learning method for medical image segmentation. By using image-level labels and clustering techniques, the paper proposes a new approach to generate pixel-level supervision to reduce over-segmentation and under-segmentation. Additionally, a self-supervised learning module is designed to complement local information in feature learning. Experimental results demonstrate the superior performance of this method in medical image segmentation and its potential for other applications.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Wei Lan, Tianchuan Yang, Qingfeng Chen, Shichao Zhang, Yi Dong, Huiyu Zhou, Yi Pan
Summary: This article proposes a novel multiview subspace clustering method, named LSGMC, to address the problems of ignoring consistent information and angular information in existing methods. LSGMC pursues a consistent low-rank structure across views and guarantees weight consistency using a symmetry constraint. It captures the inherent structure of data by utilizing fusion mechanism and employs the Schatten p-norm to obtain a low-rank coefficient matrix. Experimental results on 11 datasets demonstrate the superiority of LSGMC in clustering performance compared with ten state-of-the-art multiview clustering methods.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Geochemistry & Geophysics
Xiaoqian Zhu, Xiangrong Zhang, Tianyang Zhang, Xu Tang, Puhua Chen, Huiyu Zhou, Licheng Jiao
Summary: This article presents an intuitive and effective framework to explore the semantic and contour cues of buildings and fully excavate their correlation. By constructing an interactive dual-stream decoder, a semantic collaboration module (SCM), and a multiscale semantic context fusion module (MSCF), our method achieves superior performance in building footprint extraction.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Information Systems
Xiaozhou Lei, Zixiang Fei, Wenju Zhou, Huiyu Zhou, Minrui Fei
Summary: Low light conditions can significantly degrade image quality and cause visual task failures. Existing image enhancement technologies often suffer from overenhancement, color distortion, and high time consumption, with limited adaptability. In this study, a novel method for enhancing the lightness of low-light images is proposed. The method utilizes an energy model based on membrane vibrations induced by photon stimulations, combined with a gamma correction model. A local fusion strategy is also employed to optimize the local details of the enhanced images. Experimental results demonstrate the superiority of the proposed algorithm in terms of avoiding color distortion, restoring dark areas textures, reproducing natural colors, and reducing time cost.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Chemistry, Analytical
Ruining Yang, Xingfen Liu, Junbo Hu, Hui Xu, Jixiang Song, Huiyu Zhou, Meixing Li, Yanqin Huang, Lei Zhang, Quli Fan
Summary: A simple and robust fluorescence strategy based on a nontarget DNA-triggered catalytic hairpin assembly (CHA) is reported for improved detection of microRNA-21 in a complex matrix. The strategy achieved a wide linear range and showed significantly improved performance in diluted serum. The biosensor also demonstrated stability for miR-21 quantification, capability for analyzing single nucleotide polymorphisms (SNPs), and high-resolution imaging of miR-21 in living tumor cells, showcasing its potential for early-stage disease diagnosis and biological studies in cells.
Article
Computer Science, Artificial Intelligence
Zheheng Jiang, Zhihua Liu, Long Chen, Lei Tong, Xiangrong Zhang, Xiangyuan Lan, Danny Crookes, Ming-Hsuan Yang, Huiyu Zhou
Summary: The study proposes a novel method for tracking multiple mice and their individual parts without the need for specific tagging. The method includes using deep learning for mouse part detection and a Bayesian-inference integer linear programming model for assigning parts to individual targets. The authors also introduce a new challenging dataset and evaluate the proposed method on this dataset.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)