A5100404244- Advanced Vision and Imaging
- Human Pose and Action Recognition
- Face and Expression Recognition
- Recommender Systems and Techniques
- Medical Image Segmentation Techniques
- Advanced MRI Techniques and Applications
- Robotics and Sensor-Based Localization
- Video Surveillance and Tracking Methods
- Advanced Image and Video Retrieval Techniques
- Underwater Vehicles and Communication Systems
- Robotic Path Planning Algorithms
- Topic Modeling
- Face recognition and analysis
- Medical Imaging Techniques and Applications
- Magnetic properties of thin films
- Context-Aware Activity Recognition Systems
- Microstructure and mechanical properties
- Multimodal Machine Learning Applications
- Non-Invasive Vital Sign Monitoring
- Advanced Image Processing Techniques
- Generative Adversarial Networks and Image Synthesis
- Robot Manipulation and Learning
- nanoparticles nucleation surface interactions
- Image Enhancement Techniques
- Gait Recognition and Analysis
University of Chinese Academy of Sciences
2018-2025
Institute of Biophysics
2025
Chinese Academy of Sciences
2025
Harbin Engineering University
2023-2025
Shandong Institute of Business and Technology
2024
Multi Base (China)
2024
Beijing University of Technology
2018-2024
Zhongkai University of Agriculture and Engineering
2023
Northeast Electric Power University
2023
University of Electronic Science and Technology of China
2022
The Bluetooth-based wearable fall detection technology faces problems such as short transmission distance, easily interfered by obstacles, and high-power consumption. To address these issues, we developed a sensor board integrated with low-power ZigBee MPU6050, which can sample cache three-axial acceleration angular velocity data in the sleep mode; also designed an interrupt-driven algorithm that collect transmit to receiving end (namely server) consumption via ZigBee. Additionally, received...
Point clouds-based 3D human pose estimation that aims to recover the locations of skeleton joints plays an important role in many AR/VR applications. The success existing methods is generally built upon large scale data annotated with joints. However, it a labor-intensive and error-prone process annotate from input depth images or point clouds, due self-occlusion between body parts as well tedious annotation on clouds. Meanwhile, easier construct datasets 2D joint annotations images. To...
A dynamic distributed target hunting method is proposed for the problem of moving by multiple Autonomous Underwater Vehicles (AUVs). By integrating improved Dynamic Window Approach (DWA) with Rapidly-exploring Random Tree (RRT) algorithm and incorporating collision avoidance rules between AUVs into evaluation system DWA, are quantified, corresponding functions established. This allows selection motion trajectories that comply from predicted trajectory set, improving obstacle capability...
Abstract The hippocampus plays a critical role in memory and is prone to neural degenerative diseases. Its complex structure distinct subfields pose challenges for automatic segmentation 3 T MRI because of its limited resolution contrast. While 7 offers superior anatomical details better gray-white matter contrast, aiding clearer differentiation hippocampal structures, use restricted by high costs. To bridge this gap, algorithms synthesizing 7T-like images from scans are being developed,...
Massive amount of training samples is a challenge for 3D face recognition using deep learning frame. This paper shows method that uses twin neural network by blending depth and 2D texture. First, map generated. In order to repair holes in the model with low complexity, we those hole points into plane, then reverse them back space least square rule. Second, convolution kernel two layer channels used fuse image image. Finally, after sample pairs are generated, performed convolutional network....
This study proposes Combining Attention and Brightness Adjustment Network (CABA-Net), a deep learning network for underwater image restoration, to address the issues of color-cast, low brightness, contrast. The proposed approach achieves multi-branch ambient light estimation by extracting features different levels images achieve accurate estimates light. Additionally, an encoder-decoder transmission map module is designed combine spatial attention structures that can extract layers images’...
There are increasing real-time live applications in virtual reality, where it plays an important role capturing and retargetting 3D human pose. But is still challenging to estimate accurate pose from consumer imaging devices such as depth camera. This paper presents a novel cascaded full-body regression method single image at 100 fps. The key idea train regressors based on Gradient Boosting algorithm pre-recorded motion capture database. By incorporating hierarchical kinematics model of into...
Permanent structural changes in pure metals that are caused by plastic activity normally irreparable after unloading. Because of the lack experimental evidence, it is unclear whether can be repaired as size decreases to several nanometers; also how accommodate deformation. In this study, situ atomic-scale loading and unloading ∼2 nm Ag nanocrystals was investigated, three modes deformation were observed: (i) phase transition from face-centered cubic (fcc) hexagonal close-packed (hcp) phase,...
For the local path planning problem of autonomous underwater vehicle (AUV) in unknown complex obstacle environment, traditional Dynamic Window Approach (DWA) algorithm has problems such as easy to fall into stagnation obstacles. A based on fusion Improved DWA and Rapid-exploration Random Tree (RRT) algorithms is proposed. The improved velocity space generated according period whole dynamic window resets evaluation function; RRT plans guide points known help out state. two are fused achieve...
In practical usage, QR codes often become difficult to recognize due damage. Traditional restoration methods exhibit a limited effectiveness for severely damaged or densely encoded codes, are time-consuming, and have limitations in addressing extensive information loss. To tackle these challenges, we propose two-stage model named the EHFP-GAN, comprising an edge module code reconstruction module. The guides subsequent by repairing images, resulting finer details. hierarchical feature pyramid...
Diffusion models have achieved significant success in image and video generation. This motivates a growing interest editing tasks, where videos are edited according to provided text descriptions. However, most existing approaches only focus on for short clips rely time-consuming tuning or inference. We the first propose Video Instruction (VIDiff), unified foundation model designed wide range of tasks. These tasks encompass both understanding (such as language-guided object segmentation)...
According to the requirement of low power and accuracy for fall detection, an activity model based on three-dimensional attitude angles is introduced, difference in signal vector magnitude acceleration between daily activities falls are compared. Second, a sensor board integrated with MPU6050 ZigBee which can collect transmit tri-axial accelerations angular velocities human server at low-power developed. Finally, detection system running developed via Kalman filter kNN algorithm. It proved...
3D object tracking is the key of automatic driving. We propose a new method called Transformer-Hausdorff Net (TH-Net). It contains three main modules: Feature Extraction, Fusion, and Proposal Generation. The Extraction module extracts features from template search area, where permutation-invariable Transformers leveraged to deal with point cloud's dis-order sparsity. area are then fused by Fusion generate clues. Based on clues, we further target proposal execute verification in Generation...
In this paper, the adaptive clustering quasi-line search path planning algorithm (ACPP) is proposed from viewpoint of non-convex optimization in an artificial potential field (APF). ACPP, initial trajectory process point to target one. a complex environment, drivable area divided by edges into many relatively isolated regions, which leads problem. Therefore, based on nodes set adaptively make function each region convex field. region, environment can be perceived updating parameters with...
Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI
Time-of-flight MR angiography (TOF-MRA) allows noninvasive visualization of the mouse intracranial vessels. However, there is no available tool for vessel segmentation in 3D TOF-MRA. In this study, we developed an automated method based on a convolutional neural network and multiscale enhancement filtering algorithm. The potentially useful preclinical studies cerebrovascular diseases.
Modern MRI reconstructs images by performing Fourier transform (FT) on k-space to decompose the signal an orthogonal basis composed of trigonometric functions. We are inspired problem estimating Direction Arrival in radar theory and propose a novel route for reconstruction based multiple classification (MUSIC). MUSIC-MRI overcomes sidelobe FT significantly promotes actual resolution meaning full width at half maximum (FWHM). Our phantom experiments show that FWHMs 0.45mm 1.50mm FT, while...
Robotic grasping plays a pivotal role in real-world interactions for robots. Existing grippers often limit functionality to single mode-picking or suction. While picking handles smaller objects and suction adapts larger ones, integrating these modes breaks scale boundaries, expanding the robot's potential real applications. This article introduces transformable soft gripper capable of achieving amphibious cross-scale grasping. Despite its compact fully scalable design (20 mm diameter...
Motivation: Conventional 2D-TSE offers superior tissue contrast compared to 3D-TSE, but its slice thickness is restricted by SNR and profile of RF pulse. Goal(s): Our objective was achieve 0.4mm isotropic T2-weighted imaging with high SNR. Approach: Amplitude phase modulated gSlider pulses were developed incorporated into 2D-TSE. Their performances evaluated in phantom vivo. Results: The gSlider-TSE achieved accuracy. comparable 8 times averaged conventional image. Impact: provides...