Biography
Research Scientist @ Honda Reaserch Institute USA, Inc. focuses on SLAM and Computer Vision. Skillful in 3D Scene Reconstruction, SLAM, Object Detection & 6D Pose Estimation, Image Segmentation, Neural Radiance Fields, and Robotics Intelligent Inspection.
Download my resumé.
Interests
- Robotics
- SLAM
- 3D Scene Reconstruction
- Neural Radiance Fields
- Object 6D Pose Estimation
- Multi-View Geometry
Education
-
PhD in Electrical Engineering, 2023
The City College of New York, CUNY
-
MEng in Electrical Engineering, 2021
The City College of New York, CUNY
-
MEng in Control Engineering, 2018
Shandong University
Selected Publications
Quickly discover relevant content by filtering publications.
Robotic Inspection of Underground Utilities for Construction Survey Using a Ground Penetrating Radar
Projects
Implicit 3D Scene Reconstruction - Research Intern
Samsung Research America
- Implementing 3D indoor scene reconstruction with explicit Octree-based NeRF model andSLAM pipeline.
- Employ an explicit NeRF model with Octree structure to self-supervise tracking and mapping thread using learning approach rather than traditional visual odometry and graph optimization methods, thus we do not require trajectory ground truth to supervise network to estimate pose and mapping information.
- Deploy Intel OpenBot to automate RBG-D data collection from a mobile robotic platform
in an indoor environment.
Object Pose Estimation via RGB-D Images - Robotics Scientist Intern
ABB Robotics Group
- Generate synthetic physically based renderer images for HOPE dataset. (Partial dataset that I created has been credited by the Nvidia Learning and Perception team).
- Build up a deep network, using synthetic data to train object detector, object pose estimation.
- Experimental testing and evaluation of object pose estimation, and benchmarking with the
state-of-the-art works.
SBIR Phase-II: GPR-based Subsurface Reconstruction - Technical Leader
CCNY Robotics Lab, City College of New York
- Generate synthetic GPR dataset that covers most of the subsurface infrastructure scenarios. The
synthetic objects have different sizes, being inserted in different depth, and has multiple shapes
such as cylindrical, spherical, and cubic shape.
- Propose a novel learning-based subsurface 3D reconstruction method, that emulates the GPR-
based 3D reconstruction from the back-projection (BP) perspective and outputs a dense
volumetric map of subsurface objects.
- Develop a mathematical based 3D migration software, which use conventional back-projection
method to recover the underground targets. The software is developed on Visual Studio and
patented protect.
SBIR Phase-I: Robotic-based GPR Inspection System for Infrastructure Inspection. Co-Technical Leader
CCNY Robotics Lab, City College of New York
- Design an omnidirectional robotic GPR inspection system, this system could hold a GPR antenna to move forward, backward, sideways without any rotations, to detect the underground objects, and reconstruct the subsurface 3D objects map using GPR data. The positioning accuracy of our robotic system could reach 1 cm using Visual-Inertial Odometry and AprilTag markers.
- Develop an automated GPR data collection solution by using an omnidirectional robot, which
holds the GPR antenna on the bottom and move forward, backward and sideways without any
rotations to collect GPR data.
Robotic based 3D Migration algorithm with arbitrary motion - Research Intern
Geophysical Survey System Inc.
- This project aims to develop a 3D migration algorithm to interpret GPR data as well as design a robotic GPR data collection method.
- Develop a 3D migration method that gets rid of the limitations of conventional GPR migration
methods, that is, takes both position and orientation of GPR into consideration.
- Develop the robotic GPR data collection method instead of manually pushing the GPR cart. The
GPR data is tagged with positioning information provided by the RGB-D camera, which is
mounted on the robot.
National Nature Science Foundation: Smart and Intelligent Climbing Robot for Wind Turbine Inspection - Research Assistant
CCNY Robotics Lab, City College of New York
- Develop a tracking system for positioning and reconstruction, then navigate the robot using the map information with obstacle avoidance mission.
- Develop an impact sounding system to collect acoustic measurement data to reveal subsurface
flaws. This system includes a sliding track, solenoid, and a microphone. The solenoid is moving
along the sliding tracking while providing impacting action, and the microphone record the echo
sound.
- Focusing on processing the sounding data with Frequency Density (FD) to describe the energy of
the acoustic measurement. Furthermore, we register this FD data as well as the trajectory data on
the map built from the tracking system.
Teachings
Introduction to Robotics
Teaching Assistant, Graduate Course
Computer Engineering Lab (EE425)
Adjunct Lecturer, Undergraduate Senior Level Lab
Advanced Robotics
Teaching Assistant, Graduate Course
Contact
- jfeng1@ccny.cuny.edu
- 313-742-5794
- Steinman Hall 518, 275 Convent Ave., New York, NY 10031