A5100393714- Video Surveillance and Tracking Methods
- Infrared Target Detection Methodologies
- Fire Detection and Safety Systems
- Human Pose and Action Recognition
- Navier-Stokes equation solutions
- Advanced Chemical Sensor Technologies
- Advanced Mathematical Physics Problems
- Geometric Analysis and Curvature Flows
- Impact of Light on Environment and Health
- Advanced Image and Video Retrieval Techniques
- Advanced Image Fusion Techniques
- Infrared Thermography in Medicine
- Advanced Vision and Imaging
- Gait Recognition and Analysis
- Remote-Sensing Image Classification
- Human-Automation Interaction and Safety
- Visual Attention and Saliency Detection
- Target Tracking and Data Fusion in Sensor Networks
- Anomaly Detection Techniques and Applications
- Video Analysis and Summarization
- Advanced Differential Equations and Dynamical Systems
- Food Supply Chain Traceability
- Advanced Measurement and Detection Methods
- Human-Animal Interaction Studies
- Physical Unclonable Functions (PUFs) and Hardware Security
Chongqing Normal University
2022-2025
Harbin Institute of Technology
2016-2022
Hunan Institute of Technology
2022
Xidian University
2022
University of Technology Sydney
2022
Hunan Normal University
2014-2019
Shenzhen Institute of Information Technology
2018
Guizhou Normal University
2016-2017
North China Electric Power University
2017
National Maritime Research Institute
2006
The training of a feature extraction network typically requires abundant manually annotated samples, making this time-consuming and costly process. Accordingly, we propose an effective self-supervised learning-based tracker in deep correlation framework (named: self-SDCT). Motivated by the forward-backward tracking consistency robust tracker, multi-cycle loss as information for learning from adjacent video frames. At stage, generate pseudo-labels consecutive frames prediction under Siamese...
Existing deep Thermal InfraRed (TIR) trackers only use semantic features to represent the TIR object, which lack sufficient discriminative capacity for handling distractors. This becomes worse when feature extraction network is trained on RGB images. To address this issue, we propose a multi-level similarity model under Siamese framework robust object tracking. Specifically, compute different pattern similarities using proposed network. One of them focuses global and other computes local...
The feature models used by existing Thermal InfraRed (TIR) tracking methods are usually learned from RGB images due to the lack of a large-scale TIR image training dataset. However, these less effective in representing objects and they difficult effectively distinguish distractors because do not contain fine-grained discriminative information. To this end, we propose dual-level model containing TIR-specific correlation for robust object tracking. Specifically, inter-class objects, first...
Convolutional neural networks (CNNs) have been successfully applied to the single target tracking task in recent years. Generally, training a deep CNN model requires numerous labeled samples, and number quality of these samples directly affect representational capability trained model. However, this approach is restrictive practice, because manually labeling such large time-consuming prohibitively expensive. In article, we propose an active learning method for visual tracking, which selects...
Thermal infrared (TIR) target tracking is susceptible to occlusion and similarity interference, which obviously affects the results. To resolve this problem, we develop an Aligned Spatial-Temporal Memory network-based Tracking method (ASTMT) for TIR task. Specifically, model scene information in scenario using spatial-temporal memory network, can effectively store decrease interference of that beneficial target. In addition, use aligned matching module correct parameters network model,...
Unlike visual object tracking, thermal infrared (TIR) tracking methods can track the target of interest in poor visibility such as rain, snow, and fog, or even total darkness. This feature brings a wide range application prospects for TIR object-tracking methods. However, this field lacks unified large-scale training evaluation benchmark, which has severely hindered its development. To end, we present high-diversity single called LSOTB-TIR, consists dataset general with 1416 sequences more...
Most of the existing infrared and visible image fusion algorithms rely on hand-designed or simple convolution-based strategies. However, these methods cannot explicitly model contextual relationships between images, thereby limiting their robustness. To this end, we propose a novel Transformer-based feature network for robust that can relationship two modalities. Specifically, our consists detail self-attention module to capture information each modality saliency cross attention Since...
In this paper, we present a Large-Scale and high-diversity general Thermal InfraRed (TIR) Object Tracking Benchmark, called LSOTBTIR, which consists of an evaluation dataset training with total 1,400 TIR sequences more than 600K frames. We annotate the bounding box objects in every frame all generate over 730K boxes total. To best our knowledge, LSOTB-TIR is largest most diverse object tracking benchmark to date. evaluate tracker on different attributes, define 4 scenario attributes 12...
Existing deep Thermal InfraRed (TIR) trackers usually use the feature models of RGB for representation. However, these learned on images are neither effective in representing TIR objects nor taking fine-grained information into consideration. To this end, we develop a multi-task framework to learn TIR-specific discriminative features and correlation tracking. Specifically, first an auxiliary classification network guide generation distinguishing belonging different classes. Second, design...
Thermal infrared (TIR) pedestrian tracking is one of the important components among numerous applications computer vision, which has a major advantage: it can track pedestrians in total darkness. The ability to evaluate TIR tracker fairly, on benchmark dataset, significant for development this field. However, there not dataset. In paper, we develop dataset evaluation. includes 60 thermal sequences with manual annotations. Each sequence nine attribute labels based addition carry out...
Existing methods for video-based person re- identification (ReID) mainly learn the appearance feature of a given pedestrian via extractor and aggregator. However, models would fail to large inter-class variance when different pedestrians have similar appearances. Considering that walking postures body proportions, we propose discriminative pose beyond video retrieval. Specifically, implement two-branch architecture separately feature, then concatenate them together inference. To first detect...