Research

0. Online Temporal and Spatial Calibration for Heterogeneous Multi-Sensor Fusion 2018-to-present

  • Supervisor: Prof. Shaojie Shen ;
  • To realize accurate and robust SLAM or environment perception, sensor synchronization and extrinsic parameters calibration of different sensors are crucial to a fusion-based system, which requires sensor data alignment both temporally and spatially. Using a central IMU as the key calibration reference, all the rigidly connected sensors (IMUs, cameras, Lidars, RGBD...) in the same sensor set can be calibrated both temporally and spatially.
  • Kejie Qiu, Tong Qin, Shaojie Shen, Online Temporal and Spatial Calibration for Heterogeneous Multi-Sensor Fusion, to appear
sync_1.mp4

1. 3D human body tracking with metric scale recovery 2018-to-present

  • Supervisor: Prof. Shaojie Shen ;
  • We propose a markless 3D human tracking system using an onboard monocular visual-inertial sensor suite, without torso structure prior and scale prior of the tracked person.
  • Kejie Qiu, Wenliang Gao, Shaojie Shen, 3D Human Body Tracking With Metric Scale Recovery Using Monocular Visual-Inertial Fusion, the IEEE International Conference on Robotics and Automation, ICRA 2019, under review
icra2019HD.mp4

2. 6DoF object metric pose tracking using monocular visual-inertial fusion 2016-to-present

  • This work was supported by the WeChat-HKUST Joint Laboratory on Artificial Intelligence Technology (WHAT LAB).
  • Supervisor: Prof. Shaojie Shen ;
  • We propose a geometry-based 3D tracking system with
  • metric scale awareness for generally rigid objects using a monocular visual-inertial sensor suite, neither object scale prior nor specific motion assumption of the dynamic object is needed.
  • We complete the 3D tracking system by estimating the metric scale of the tracked object based on trace correlation analysis and observability condition check.
  • We implement the 3D tracking system to verify the tracking performance in comparison with ground truth and demonstrate the AR application potential.
  • Publication:
  • Kejie Qiu, Tong Qin, Wenliang Gao, Shaojie Shen, Tracking 3 D Motion of Dynamic Objects Using Monocular Visual-Inertial Sensing, IEEE Transactions on Robotics, TRO, accepted
  • Kejie Qiu, Tong Qin, Hongwen Xie, Shaojie Shen, Estimating metric poses of dynamic objects using monocular visual-inertial fusion, the IEEE International Conference on Intelligent Robots and Systems, IROS 2018
iros2018kejie_final.mp4

3. Model-based global localization for aerial robots 2015-to-2016

  • This work was supported by the WeChat-HKUST Joint Laboratory on Artificial Intelligence Technology (WHAT LAB).
  • Supervisor: Prof. Shaojie Shen ;
  • We propose a practical global localization system for aerial robots operating in large-scale urban environments by utilizing offline-generated 3D models.
  • We design block rendering scheme, electronic image stabilization system, and edge-based image-to-model registration method, to achieve real-time and robust pose
  • estimation.
  • We implement the complete autonomous aerial robot system platform to verify localization performance and demonstrate closed-loop control capability.
  • Publication:
  • Kejie Qiu, Tianbo Liu, Shaojie Shen, Model-based global localization for aerial robots using edge alignment, IEEE Robotics and Automation Letters, RAL, with ICRA 2017 presentation
  • Kejie Qiu, Shaojie Shen, Model-aided monocular visual-inertial state estimation and dense mapping, the IEEE International Conference on Intelligent Robots and Systems, IROS 2017
ral2016kejie.mp4

4. Visible Light Communication based indoor localization and path-planning system Sep, 2013-to-2015

  • This work was supported by the HKUST Project IGN13EG03; partially sponsored by the Research Grant Council of Hong Kong SAR Government, China, under project No. 16206014 and NSFC Fund 6140021318.
  • Supervisor: Prof. Ming Liu ; Cooperator: Fangyi Zhang.
  • Developed a beacon code selection algorithm and corresponding signal decomposition algorithm based on Gold-sequence and its correlation properties.
  • Developed a light intensity distribution model based on Gaussian Process Regression (GPR).
  • Developed a localization algorithm using Bayesian filter based on the GPR model.
  • Publication:
  • Kejie Qiu, Fangyi Zhang, Ming Liu, Let the light guide us: VlC-based localization, IEEE Robotics & Automation Magazine, RAM
  • Kejie Qiu, Fangyi Zhang, Ming Liu, Visible light communication-based indoor localization using Gaussian Process, the IEEE International Conference on In-

telligent Robots and Systems, IROS 2015

  • Kejie Qiu, Fangyi Zhang, Ming Liu, Visible light communication-based indoor environment modeling and metric-free path planning, the IEEE International

Conference on Automation Science and Engineering, CASE 2015

  • Ming Liu, Kejie Qiu, F Che, S Li, B Hussain, L Wu, CP Yue, Towards indoor localization using visible light communication for consumer electronic devices, the

IEEE International Conference on Intelligent Robots and Systems, IROS 2014

5. Vision Navigation based Unmanned Smart Car(3rd national university opt-sci-tech competition) Mar 2012 - Aug 2012

  • We designed and made a racing car with self-navigation system using a digital camera, a microprocessor and a variety of peripheral components. Our vision based solution performed pretty well in the keen 3rd national university students’ opt-sci-tech competition. We won the second prize award in the smart-car group.
  • The grid-shape track used for the time trial was defined by pairs of red lines. The self-navigation car should run along the track without touching the red lines, make right decisions(turn right, turn left or go straight) when reaching crossings according to the randomly defined turning markers, and finally drive itself into the garage without touching red lines either since touching will invalidate the racing record. The car finishes all these operations in minimum time will win.

6. Steel Plate thickness Measurement System based on Lasers and PSD component May 2011 - May 2012

  • Principle investigator of this Student Research Training Project(SRTP)
  • The project focuses on a steel plate thickness detection system based on lasers and Position Sensitive Detector(PSD) component. The system consists of optical module, signal processing module and computing module. A series of experiments have been designed and conducted with the help of Professor Haijun Zhang, we construct the optical system, electric circuit, and a C++ based application client in the end.

7. Analysis of Solar transmission loss in the Solar-Thermal Power System Nov 2011 - Jan 2013

  • Principle investigator of this ‘SUPCON’ Educational Fund project
  • Focused on analysis of solar light efficiency in the Solar-Thermal Power System with financial support of 10k RMB. With the assist of Professor Zhaofeng Cen and Advanced Engineer Wenjun Huang, we finished the model construction and the simulation analysis of the light spot in the optical system using Tracepro and MATLAB respectively.