Wei Wu

Northwest Institute of Nuclear Technology

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Xiangqin Zhu

Northwest Institute of Nuclear Technology

Haitao Gan

Hubei University of Technology

Cewen Liu

Tsinghua University

Wei Jia

Northwest Institute of Nuclear Technology

Mingjie Guo

China National Petroleum Corporation (China)

Haohuan Fu

Tsinghua University

Nanxun Dai

China National Petroleum Corporation (China)

Fan Guo

Northwest Institute of Nuclear Technology

Qiang Du

Lawrence Berkeley National Laboratory

Niki Trigoni

University of Oxford

Cooperative Institutions

Institute of High Energy Physics

241

Chinese Academy of Sciences

114

China Center of Advanced Science and Technology

Tsinghua University

University of Chinese Academy of Sciences

COMSATS University Islamabad

University of Science and Technology of China

Beihang University

Central China Normal University

Sichuan University

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Selective Sensor Fusion for Neural Visual-Inertial Odometry

2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (2019)

Changhao Chen Stefano Rosa Yishu Miao Chris Xiaoxuan Lu Wei Wu

Deep learning approaches for Visual-Inertial Odometry (VIO) have proven successful, but they rarely focus on incorporating robust fusion strategies for dealing with imperfect input sensory data. We propose a novel end-to-end selective sensor fusion framework for monocular VIO, which fuses monocular images and inertial measurements in order to estimate the trajectory whilst improving robustness to real-life issues, such as missing and corrupted data or bad sensor synchronization. In particular, we propose two fusion modalities based on different masking strategies: deterministic soft fusion and stochastic hard fusion, and we compare with previously proposed direct fusion baselines. During testing, the network is able to selectively process the features of the available sensor modalities and produce a trajectory at scale. We present a thorough investigation on the performances on three public autonomous driving, Micro Aerial Vehicle (MAV) and hand-held VIO datasets. The results demonstrate the effectiveness of the fusion strategies, which offer better performances compared to direct fusion, particularly in presence of corrupted data. In addition, we study the interpretability of the fusion networks by visualising the masking layers in different scenarios and with varying data corruption, revealing interesting correlations between the fusion networks and imperfect sensory input data.

Odometry

Sensor Fusion

Robustness

Interpretability

Visual Odometry

Monocular

10.1109/cvpr.2019.01079

Cite

Citations (132)

Realization of a Neural Network Parallel MIMD Processor

Dianzi Ke-ji Daxue xuebao (2008)

Wei Wu

To expand the realtime and embedded application of the neural network in industry control,general purpose neural netowrk processor(APP) based on multiple instruction stream and multiple data stream(MIMD) is paoposed. The tasks among PEGs,PEG,and MAC coprocessor can be processed in parallel,and also communications among PEGs and other memorys can be carried out in paralled. The processor has been simulated on FPGA and is used to implement two different neural network:the BP network used in roll eccentricity online control and the Hopfield network used in character recogonition. The simulation result shows that the performance of the proposed APP is better than that of other general methods for neural network implementition.

MIMD

Coprocessor

Realization (probability)

Network processor

Source

Cite

Citations (0)

Selective Sensor Fusion for Neural Visual-Inertial Odometry

arXiv (Cornell University) (2019)

Changhao Chen Stefano Rosa Yishu Miao Chris Xiaoxuan Lu Wei Wu

Odometry

Sensor Fusion

Interpretability

Robustness

Visual Odometry

Monocular

10.48550/arxiv.1903.01534

Cite

Citations (7)

Calculation of coaxial line X-ray responses by transmission line method and finite difference time domain method

High Power Laser and Particle Beams (2007)

Wei Wu

For calculating the coaxial line responses in X rays environment,transmission line(TL) method and finite difference time domain(FDTD) method are presented up respectively.Using charges of the central wire,distributed charge in insulation and the result of test,the current source of TL equations and the current density of Maxwell equations are presented.As an example,the central wire currents of the load under the condition of the matching load of cable ends are calculated using the two methods for energy of 60 keV,and the central wire currents of different energies are calculated with TL method.Numerical results indicate that the responses currents of the two methods are consistent,electrostatic field exists in the insulator,and the central wire currents have a maximum value for different energies.Photoelectric parameters are calculated using Monte-Carlo method.

Coaxial

Telegrapher's equations

Photoelectric effect

Source

Cite

Citations (1)

Study on the Parameters of Cavity System-generate Electromagnetic Pulse (SGEMP) Calculation Model Base on Maxwell-Valsov Equation

2020 IEEE 3rd International Conference on Automation, Electronics and Electrical Engineering (AUTEEE) (2020)

Wei Wu Mo Zhao Jinxi Li Hui Zhou

In this paper, a calculation model based on Maxwell equation and Vlasov equation is established to solve the electromagnetic pulse problem in a cavity radiated by X-ray. When solving the common cavity model, the choice of geometric space step is mainly considered as the constraint of calculation resources and the thickness of space charge limiting layer. The velocity space grid is recommended to be divided into 11 segments. If the calculation amount is too large, the velocity space could be reduced to 9 or 7 segments, but the calculation accuracy will be slightly lower.

Electromagnetic pulse

Limiting

Plasma Modeling

Base (topology)

10.1109/auteee50969.2020.9315620

Cite

Citations (1)

On the Design of Incremental Predictive Functional Control Algorithm

Lecture notes in electrical engineering (2012)

Wei Wu Qiang Yang Wenjun Yan

Robustness

Model Predictive Control

Internal model

10.1007/978-1-4471-2467-2_2

Cite

Citations (0)

Analysis on Dynamic Change and Tendency of Land-cover based on Object\|oriented Classification

Yaogan jishu yu yingyong (2014)

Tianjun Wu Xiao Hu Xia Liegang Jiancheng Luo Zhanfeng Shen

Land Cover

Object based

Change analysis

10.11873/j.issn.1004-0323.2014.4.0600

Cite

Citations (0)

Fabrication of Sodium Alginate Porous Scaffolds for Heavy Metal Ion Removal from Aqueous Solution

Journal of Macromolecular Science Part B (2022)

Mengxiang Zeng Wei Wu Jianjun Fang Zhihua Zhou Ying Lan

The adsorption method is recognized as an effective technology for dealing with water pollution caused by heavy metal ions. For a highly efficient, eco-friendly and low-cost adsorption process, the adsorbent is the deciding factor. In this study a porous sodium alginate scaffold was prepared by the method of freeze-drying as a heavy metal ion adsorbent. The parameters affecting the morphology and adsorption characteristics of the scaffolds were investigated. The results showed that sodium alginate porous scaffolds could be prepared under the concentration in the range of 1–5%. After calcium ion pre-crosslinking, the sodium alginate scaffold was used to adsorb the heavy metal ions of Cd2+, Pb2+, and Cu2+ from aqueous solution and the maximum adsorption efficiency under optimized adsorption conditions were 99.43%, 73.67%, and 59.25%, respectively. At the same time, the scaffold could be used for desorption and recycling. After five cycles, the removal percentages of Pb2+, Cd2+, and Cu2+ by the sodium alginate scaffold decreased by 25%, 25%, and 16%, respectively.

Sodium alginate

Calcium alginate

10.1080/00222348.2023.2175513

Cite

Citations (3)

Regularity analysis of leakage-field from vertically polarized bounded wave electromagnetic pulse simulator with distributed load

High Power Laser and Particle Beams (2020)

Linshen Xie Wei Wu Zhu Xiang-qin

A parallel finite difference time domain (FDTD) method combined with perfectly magnetic conductor (PMC) image method and CST software is firstly presented for computing the leakage field ( including sideward and backward leakage field) from vertically polarized bounded wave electromagnetic pulse (EMP) simulator with distributed load. The results got from simulation agree well with those from experiment. Numerical results show that the leakage field’s peak-value near the ground reaches maximum in the direction of height, but there’s little difference of leakage fields’ peak-values at the testing points under 1.5 m height and far away from the simulator; The peak-values of sideward leakage field from simulator’s transitional section are larger than those from distributed load section, both of them are larger than the backward leakage field from the end of distributed load section, as the testing point is near the simulator, and the backward leakage field may be larger than those sideward leakage fields, as the vertically distance between the testing point and the simulator increases; The regularity conclusion suits for the simulator with different double exponential source. The more excited source component in high frequency domain, the larger backward leakage field in some distance domain. The leakage field from the simulator increases, as ground’s conductivity increases.

Leakage (economics)

Electromagnetic pulse

10.11884/hplpb202032.190434

Cite

Citations (1)

Dual comb generation in triple normal dispersion coupled microresonators using avoided mode crossing

Wei Wu Dongmei Huang P. K. A. Wai

Repetition-rate-synchronized dual-combs are generated in triple coupled microresonators operating in the normal dispersion by exploiting avoided mode crossing, opening up new possibilities for dual-comb generation and integrated applications.

Dual mode

Mode (computer interface)

10.1364/cleo_si.2024.sw3f.5

Cite

Citations (0)