A5019633896- Gait Recognition and Analysis
- Advanced Memory and Neural Computing
- Human Pose and Action Recognition
- Video Surveillance and Tracking Methods
- Radiation Effects in Electronics
- Robotics and Sensor-Based Localization
- Advanced Vision and Imaging
- Modular Robots and Swarm Intelligence
- Molecular Communication and Nanonetworks
- Optimization and Search Problems
- Hand Gesture Recognition Systems
- Robotic Path Planning Algorithms
- Image Processing Techniques and Applications
- Injury Epidemiology and Prevention
- Advanced Image Processing Techniques
- Diabetic Foot Ulcer Assessment and Management
- Underwater Vehicles and Communication Systems
- Advanced Image and Video Retrieval Techniques
- Teleoperation and Haptic Systems
- Target Tracking and Data Fusion in Sensor Networks
- Face recognition and analysis
- Robotic Locomotion and Control
- Guidance and Control Systems
- Gaze Tracking and Assistive Technology
- Neuroscience and Neural Engineering
University of Maryland, College Park
2023-2025
Zhejiang University
2022-2024
China Mobile (China)
2023
Nanjing Institute of Technology
2021-2022
Fiberhome Technology Group (China)
2021-2022
Huazhong University of Science and Technology
2015-2022
Huzhou University
2022
Ningbo Institute of Industrial Technology
2021
State Key Laboratory of Industrial Control Technology
2021
Although gait recognition has drawn increasing research attention recently, it remains challenging to learn discriminative temporal representation since the silhouette differences are quite subtle in spatial domain. Inspired by observation that humans can distinguish gaits of different subjects adaptively focusing on sequences with time scales, we propose a context-sensitive feature learning (CSTL) network this paper, which aggregates features three scales obtain motion according contextual...
Neuromorphic vision sensors or event cameras have made the visual perception of extremely low reaction time possible, opening new avenues for high-dynamic robotics applications. These cameras’ output is dependent on both motion and texture. However, camera fails to capture object edges that are parallel motion. This a problem intrinsic sensor therefore challenging solve algorithmically. Human deals with perceptual fading using active mechanism small involuntary eye movements, most prominent...
This paper presents a parallax-robust video stitching technique for timely synchronized surveillance video. An efficient two-stage procedure is proposed in this to build wide Field-of-View (FOV) videos applications. In the model calculation stage, we develop layered warping algorithm align background scenes, which location-dependent and turned out be more robust parallax than traditional global projective methods. On selective seam updating propose change-detection based optimal selection...
The development of aerial autonomy has enabled robots to fly agilely in complex environments. However, dodging fast-moving objects flight remains a challenge, limiting the further application unmanned vehicles (UAVs). bottleneck solving this problem is accurate perception rapid dynamic objects. Recently, event cameras have shown great potential problem. This paper presents complete system including ego-motion compensation, object detection, and trajectory prediction for with low latency high...
Gaze isan intuitive and direct way to represent the intentions of an individual. However, when it comes assistive aerial teleoperation which aims perform operators' intention, rare attention has been paid gaze. Existing methods obtain intention directly from remote controller (RC) input, is unfriendly non-professional operators, as experimental results show in <xref ref-type="sec" rid="sec6c" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Section...
Gait recognition is a promising long-distance biological technology that widely used in public security and video surveillance. In the most recent literature, researchers use fixed horizontal partitions on input images or intermediate features for feature extraction, which are neither adaptive to variations camera viewpoint individual appearances nor sufficient spatio-temporal fine-grained capturing. Besides, existing methods either ignore temporal modeling model...
Graph convolutional networks have been widely applied in skeleton-based gait recognition. A key challenge this task is to distinguish the individual walking styles of different subjects across various views. Existing state-of-the-art methods employ uniform convolutions extract features from diverse sequences and ignore effects viewpoint changes. To overcome these limitations, we propose a condition-adaptive graph (CAG) convolution network that can dynamically adapt specific attributes each...
The rapidly increasing capabilities of autonomous mobile robots promise to make them ubiquitous in the coming decade. These will continue enhance efficiency and safety novel applications such as disaster management, environmental monitoring, bridge inspection, agricultural inspection. To operate autonomously without constant human intervention, even remote or hazardous areas, must sense, process, interpret data using only onboard sensing computation. This capability is made possible by...
Multicopters are able to perform high maneuverability yet their potential have not been fully achieved. In this work, we propose a full-body, optimization-based motion planning framework that takes shape and attitude of aerial robot into consideration such the aggressiveness drone maneuvering improves significantly in cluttered environment. Our method series intersecting polyhedrons describe range 3D free spaces outputs time-indexed trajectory real-time with full-body collision-free...
Person reidentification (re-id) has been widely studied because of its extensive use in video surveillance and forensics applications. It aims to search a specific person among nonoverlapping camera network, which is highly challenging due large variations the cluttered background, human pose, viewpoint. We present metric learning algorithm for Mahalanobis distance re-id. Generally speaking, there exist two forces conventional process, one pulling force that pulls points same class closer...
Neuromorphic vision sensors or event cameras have made the visual perception of extremely low reaction time possible, opening new avenues for high-dynamic robotics applications. These cameras' output is dependent on both motion and texture. However, camera fails to capture object edges that are parallel motion. This a problem intrinsic sensor therefore challenging solve algorithmically. Human deals with perceptual fading using active mechanism small involuntary eye movements, most prominent...
One of the core activities an active observer involves moving to secure a "better" view scene, where definition is task-dependent. This paper focuses on task human pose estimation from videos capturing person's activity. Self-occlusions within scene can complicate or even prevent accurate estimation. To address this, relocating camera new vantage point necessary clarify view, thereby improving 2D formalizes process achieving improved viewpoint. Our proposed solution this challenge comprises...
This paper presents a low-cost, centralized modular underwater robot platform, ModCube, which can be used to study swarm coordination for wide range of tasks in environments. A ModCube structure consists multiple robots. Each move six DoF with eight thrusters and rigidly connected other robots an electromagnet controlled by onboard computer. In this paper, we present novel method characterizing visualizing dynamic behavior, along four benchmarks evaluate the morphological performance robot....
When observing objects, humans benefit from their spatial visualization and mental rotation ability to envision potential optimal viewpoints based on the current observation. This capability is crucial for enabling robots achieve efficient robust scene perception during operation, as provide essential informative features accurately representing scenes in 2D images, thereby enhancing downstream tasks. To endow with this human-like active viewpoint optimization capability, we propose...
This paper investigates Path planning Among Movable Obstacles (PAMO), which seeks a minimum cost collision-free path among static obstacles from start to goal while allowing the robot push away movable (i.e., objects) along its when needed. To develop planners that are complete and optimal for PAMO, planner has search giant state space involving both location of as well locations objects, grows exponentially with respect number objects. The main idea in this is that, only small fraction...
Margin-based metric losses have shown great success in Person Re-identification and Face Verification. But most existing works adopt a fixed class-level margin regardless of the difference between each training sample. This paper proposes Norm-Aware Margin Assignment (NAMA) scheme to dynamically adjust weight sample during training. Combined with margin-based classification losses, NAMA improves robustness feature embedding by assigning larger margins more recognizable samples. is fully...