A5038350490- Robot Manipulation and Learning
- Robotics and Sensor-Based Localization
- Image Processing Techniques and Applications
- Human Pose and Action Recognition
- Hand Gesture Recognition Systems
- Image and Object Detection Techniques
- Industrial Vision Systems and Defect Detection
- Advanced Vision and Imaging
- Advanced Neural Network Applications
- Soft Robotics and Applications
- 3D Shape Modeling and Analysis
- EEG and Brain-Computer Interfaces
- Indoor and Outdoor Localization Technologies
- Tactile and Sensory Interactions
- Autonomous Vehicle Technology and Safety
- Advanced Sensor and Energy Harvesting Materials
- Sparse and Compressive Sensing Techniques
- Integrated Circuits and Semiconductor Failure Analysis
Hunan University
2018-2025
Robotic grasping is one of the key functions for realizing industrial automation and human–machine interaction. However, current robotic methods unknown objects mainly focus on generating 6-D grasp poses, which cannot obtain rich object pose information are not robust in challenging scenes. Based this, this article, we propose a continuous system that achieves end-to-end intraclass 3-D space by accurate category-level estimation. Specifically, to achieve estimation, first, global shape...
Tracking the 6-degree-of-freedom (6D) object pose in video sequences is gaining attention because it has a wide application multimedia and robotic manipulation. However, current methods often perform poorly challenging scenes, such as incorrect initial pose, sudden re-orientation, severe occlusion. In contrast, we present robust 6D tracking method with novel hierarchical feature fusion network, refer HFF6D, which aims to predict object's relative between adjacent frames. Instead of...
Six-degree-of-freedom (6DoF) object pose estimation is a crucial task for virtual reality and accurate robotic manipulation. Category-level 6DoF has recently become popular as it improves generalization to complete category of objects. However, current methods focus on data-driven differential learning, which makes them highly dependent the quality real-world labeled data limits their ability generalize unseen To address this problem, we propose multi-hypothesis (MH) consistency learning...
Nine-degrees-of-freedom (9-DoF) object pose and size estimation is crucial for enabling augmented reality robotic manipulation. Category-level methods have received extensive research attention due to their potential generalization intra-class unknown objects. However, these require manual collection labeling of large-scale real-world training data. To address this problem, we introduce a diffusion-based paradigm domain-generalized category-level 9-DoF estimation. Our motivation leverage the...
Nine-degrees-of-freedom (9-DoF) object pose and size estimation is crucial for enabling augmented reality robotic manipulation. Category-level methods have received extensive research attention due to their potential generalization intra-class unknown objects. However, these require manual collection labeling of large-scale real-world training data. To address this problem, we introduce a diffusion-based paradigm domain-generalized category-level 9-DoF estimation. Our motivation leverage the...
Currently, most mobile devices have WIFI and camera modules to locate their position. However, there are two main challenges in large, highly similar indoor environments (localization accuracy localization time). Aiming balance these problems, we propose a sequential-multi-decision integrated system that combines vision acquire users' locations. This has phases: sequential fusion adaptive multi-decision localization. The former employs WIFI-based first, then image-based used within the...
Vision-guided robotic picking in 3-D space is a key technology for industrial automation and intelligent manufacturing. However, existing methods rely on labeled real-world data learning, significantly limiting their ability to generalize novel objects robustness challenging scenes containing occlusions clutter. To address these problems, we propose domain-generalized method (DGPF6D) that builds contrastive learning-based 6-D pose estimation. DGPF6D generalizes by training only synthetic...
Object pose estimation is a fundamental computer vision problem with broad applications in augmented reality and robotics. Over the past decade, deep learning models, due to their superior accuracy robustness, have increasingly supplanted conventional algorithms reliant on engineered point pair features. Nevertheless, several challenges persist contemporary methods, including dependency labeled training data, model compactness, robustness under challenging conditions, ability generalize...
We propose a real-time detection method for vehicle queue length at intersections based on image processing technology. Firstly, we acquire the of intersection, and apply automatic brightness adjustment algorithm lane line to reduce affect different light intensity camera shake respectively. And then, preprocessed is subtracted from background obtain foreground vehicle. Finally, detect by middle line, actual measured calibration method. The experimental results show that proposed has high...
Estimating the 6D object pose has attracted a quantity of research attention because it is significant for intelligence manufacturing, autonomous driving, and virtual reality. Previous works have focused on using complex network structures to improve accuracy estimation, ignoring real-time performance. In this paper, lightweight effective proposed high-precision category-level estimation based depth image. Specifically, our three steps. Firstly, we use PointNet++ extract geometric features...
Monocular 6D object pose estimation aims to estimate 6 degrees of freedom known objects, gaining attention. Correspondence-based methods are the mainstream methods. They analyze geometric information in 2D RGB images and establish 2D-3D correspondences calculate pose. However, accuracy suffers from that can not provide enough information. To solve this problem, We propose a novel prior geometry guided direct regression network (PGDRN), which fully uses knowledge contained given models....
Estimating the 6D pose of known objects has attracted a lot research attention since it is important for intelligent robot manipulation and virtual reality. In this paper, in order to improve performance estimation using RGB image, we propose novel object framework based on monocular regression depth. To get depth map use U-Net regress information effectively. For estimation, our approach two steps, which first uses Convolutional Neural Network (CNN) extract feature data regressed data,...