A5100699151- Robotics and Sensor-Based Localization
- 3D Surveying and Cultural Heritage
- Advanced Vision and Imaging
- Video Surveillance and Tracking Methods
- Indoor and Outdoor Localization Technologies
- Advanced Image and Video Retrieval Techniques
- Advanced Neural Network Applications
- 3D Shape Modeling and Analysis
- Robotic Path Planning Algorithms
- Traffic Prediction and Management Techniques
- Remote Sensing and LiDAR Applications
- Transportation Planning and Optimization
- BIM and Construction Integration
- Augmented Reality Applications
- Human Pose and Action Recognition
- Advanced Computational Techniques and Applications
- Image and Object Detection Techniques
- Computer Graphics and Visualization Techniques
- Interactive and Immersive Displays
- Advanced Algorithms and Applications
- Emotion and Mood Recognition
- Advanced Software Engineering Methodologies
- Modular Robots and Swarm Intelligence
- Innovations in Concrete and Construction Materials
- Advanced Sensor and Control Systems
Wuhan University
2004-2025
Commercial Aircraft Corporation of China (China)
2025
Southwest University
2022-2025
Chongqing University of Posts and Telecommunications
2018-2025
New York University
2018-2024
Hong Kong University of Science and Technology
2023-2024
University of Hong Kong
2023-2024
Xiamen University
2024
Nanjing University of Finance and Economics
2024
Shibaura Institute of Technology
2018-2024
Recent deep networks that directly handle points in a point set, e.g., PointNet, have been state-of-the-art for supervised learning tasks on clouds such as classification and segmentation. In this work, novel end-to-end auto-encoder is proposed to address unsupervised challenges clouds. On the encoder side, graph-based enhancement enforced promote local structures top of PointNet. Then, folding-based decoder deforms canonical 2D grid onto underlying 3D object surface cloud, achieving low...
The recent growing interest for indoor Location-Based Services (LBSs) has created a need more accurate and real-time positioning solutions. sparse nature of location finding makes the theory Compressive Sensing (CS) desirable using Received Signal Strength (RSS) from Wireless Local Area Network (WLAN) Access Points (APs). We propose an RSS-based system compressive sensing, which is method to recover signals small number noisy measurements by solving `1-minimization problem. Our estimator...
Unlike on images, semantic learning 3D point clouds using a deep network is challenging due to the naturally unordered data structure. Among existing works, PointNet has achieved promising results by directly sets. However, it does not take full advantage of point's local neighborhood that contains fine-grained structural information which turns out be helpful towards better learning. In this regard, we present two new operations improve with more efficient exploitation structures. The first...
It is challenging to measure the geometry distortion of point cloud introduced by compression. Conventionally, errors between clouds are measured in terms point-to-point or point-to-surface distances, that either ignores surface structures heavily tends rely on specific reconstructions. To overcome these drawbacks, we propose using point-to-plane distances as a geometric distortions The intrinsic resolution proposed normalizer convert mean square PSNR numbers. In addition, perceived local...
Boundary and edge cues are highly beneficial in improving a wide variety of vision tasks such as semantic segmentation, object recognition, stereo, proposal generation. Recently, the problem detection has been revisited significant progress made with deep learning. While classical is challenging binary itself, category-aware by nature an even more multi-label problem. We model that each pixel can be associated than one class they appear contours or junctions belonging to two classes. To this...
We present a simultaneous localization and mapping (SLAM) algorithm for hand-held 3D sensor that uses both points planes as primitives. show it is possible to register data in two different coordinate systems using any combination of three point/plane primitives (3 planes, 2 1 point, plane points, 3 points). Our the minimal set RANSAC framework robustly compute correspondences estimate pose. As number significantly smaller than typical data, our prefers primitive combinations involving more...
Real-time plane extraction in 3D point clouds is crucial to many robotics applications. We present a novel algorithm for reliably detecting multiple planes real time organized obtained from devices such as Kinect sensors. By uniformly dividing cloud into non-overlapping groups of points the image space, we first construct graph whose node and edge represent group their neighborhood respectively. then perform an agglomerative hierarchical clustering on this systematically merge nodes...
Abstract Existing multispectral imagers mostly use available array sensors to separately measure 2D data slices in a 3D spatial-spectral cube. Thus they suffer from low photon efficiency, limited spectrum range and high cost. To address these issues, we propose conduct imaging using single bucket detector, take full advantage of its sensitivity, wide range, cost, small size light weight. Technically, utilizing the detector’s fast response, scene’s information is multiplexed into dense 1D...
Short-term passenger flow forecasting is an essential component in urban rail transit operation. Emerging deep learning models provide good insight into improving prediction precision. Therefore, we propose a architecture combining the residual network (ResNet), graph convolutional (GCN), and long short-term memory (LSTM) (called "ResLSTM") to forecast on scale. First, improved methodologies of ResNet, GCN, attention LSTM are presented. Then, model proposed, wherein ResNet used capture...
We present a review of 3D point cloud processing and learning for autonomous driving. As one the most important sensors in vehicles (AVs), lidar collect clouds that precisely record external surfaces objects scenes. The tools are critical to map creation, localization, perception modules an AV. Although much attention has been paid data collected from cameras, such as images videos, increasing number researchers have recognized importance significance driving proposed algorithms exploit...
Vehicle-to-everything (V2X) communication techniques enable the collaboration between vehicles and many other entities in neighboring environment, which could fundamentally improve perception system for autonomous driving. However, lack of a public dataset significantly restricts research progress collaborative perception. To fill this gap, we present V2X-Sim, comprehensive simulated multi-agent V2X-aided V2X-Sim provides: (1) sensor recordings from road-side unit (RSU) multiple that...
Abstract Antitumor therapies based on adoptively transferred T cells or oncolytic viruses have made significant progress in recent years, but the limited efficiency of their infiltration into solid tumors makes it difficult to achieve desired antitumor effects when used alone. In this study, an virus (rVSV-LCMVG) that is not prone induce virus-neutralizing antibodies was designed and combined with cells. By transforming immunosuppressive tumor microenvironment immunosensitive one, B16...
An indoor tracking and navigation system based on measurements of received signal strength (RSS) in wireless local area network (WLAN) is proposed. In the system, location determination problem solved by first applying a proximity constraint to limit distance between coarse estimate current position previous estimate. Then, Compressive Sensing-based (CS--based) positioning scheme, proposed our work , applied obtain refined The used with map-adaptive Kalman filter, which assumes linear motion...
To reduce cost in storing, processing and visualizing a large-scale point cloud, we consider randomized resampling strategy to select representative subset of points while preserving application-dependent features.The proposed is based on graphs, which can represent underlying surfaces lend themselves well efficient computation.We use general feature-extraction operator features propose reconstruction error evaluate the quality resampling.We obtain form optimal distribution by minimizing...
Automatic detection and classification of defects in infrastructure surface images can largely boost its maintenance efficiency. Given enough labeled images, various supervised learning methods have been investigated for this task, including decision trees support vector machines previous studies, deep neural networks more recently. However, real world applications, labels are harder to obtain than due the limited labeling resources (i.e., experts). Thus we propose a active system maximize...
The sparse nature of location finding problem makes the theory compressive sensing desirable for indoor positioning in Wireless Local Area Networks (WLANs). In this paper, we address received signal strength (RSS)-based localization WLANs using (CS), which offers accurate recovery signals from a small number measurements by solving an ¿ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -minimization problem. A pre-processing procedure...
Modern face alignment methods have become quite accurate at predicting the locations of facial landmarks, but they do not typically estimate uncertainty their predicted nor predict whether landmarks are visible. In this paper, we present a novel framework for jointly landmark locations, associated uncertainties these and visibilities. We model as mixed random variables them using deep network trained our proposed Location, Uncertainty, Visibility Likelihood (LUVLi) loss. addition, release an...
Real-time road traffic congestion monitoring is an important and challenging problem. Most existing approaches require the deployment of infrastructure sensors or large-scale probe vehicles. Their installation often expensive temporal-spatial coverage limited. Probe vehicle data are oftentimes noisy on urban arterials, therefore insufficient to provide accurate estimation. This paper presents a novel social-media based approach monitoring, in which pedestrians, drivers, passengers retreated...
In the domain of visual tracking, most deep learning-based trackers highlight accuracy but casting aside efficiency. Therefore, their real-world deployment on mobile platforms like unmanned aerial vehicle (UAV) is impeded. this work, a novel two-stage Siamese network-based method proposed for i.e., stage-1 high-quality anchor proposal generation, stage-2 refining proposal. Different from anchor-based methods with numerous pre-defined fixed-sized anchors, our no-prior can 1) increase...