ICCV paper accepted
September 2013

Our paper (co-author for me) about the PatchMatch Stereo algorithm has been accepted for poster presentation at ICCV in Sydney

IROS paper accepted
June 2013

Our paper about efficient compositional approached for direct visual odometry, has been accepted for oral presentation at the IEEE IROS Conference in Tokyo, Japan

Research interests TUM Robotics and Embedded Systems

Robot Vision
Localization, Place Recognition, Simultaneous Localization and Mapping (SLAM), Visual Servoing, Sensor Fusion
Computer Vision
VSLAM (Mono, Stereo, RGB-D), Sparse Bundle Adjustment, Visual Tracking, Detection, Features, Image Alignment, 3D Reconstruction, Direct/Dense Methods for Visual Odometry
Embedded and Parallel computing
Computer Vision on Multicore ARM, GP-GPU Programming (OpenCL), Hardware Acceleration (e.g. OpenCL on FPGA)

Efficient Compositional Approaches for Real-Time Robust Direct Visual Odometry from RGB-D Data
IEEE/RSJ International Conference on Intelligent Robots and Systems, November 2013

In this paper we give an evaluation of different methods for computing frame-to-frame motion estimates for a moving RGB-D sensor, by means of aligning two images using photometric error minimization. These kind of algorithms have recently shown to be very accurate and robust and therefore provide an attractive solution for robot ego-motion estimation and navigation. We demonstrate three different alignment strategies, namely the Forward-Compositional, the Inverse-Compositional and the Efficient Second-Order Minimization approach, in a general robust estimation framework. We further show how estimating global affine illumination changes, in general improves the performance of the algorithms. We compare our results with recently published work, considered as state-of-the art in this field, and show that our solutions are in general more precise and can perform in real-time on standard hardware.

Project Page at TU München

More to come ...

Markerless, Vision-Assisted Flight Control of a Quadrocopter
IEEE/RSJ International Conference on Intelligent Robots and Systems, October 2010

Abstract— In this paper, we present a system for controlling a quadrocopter using both optical and inertial measurements. We show how to use external stereo camera measurements for visual servoing, by onboard fusion at high rates, only natural features provided by the vehicle and without any active marker. In our experiments, we show the accuracy and robustness of our system during indoor flights, as well as robustness to external flight disturbances.
Associated Video:

More Videos

For more have a look at the IAFC project page

Diploma Thesis
2D/3D Hand Detection and Tracking

2D frame-by-frame hand detection

2D tracking using lines and color statistics (CCD)

3D using multiview setup and edge information

Student Project 2006

Implementation of ICondensation particle filter for 2D tracking
Condensation with background clutter

Condensation tennis-ball tracking

ICondensation in room

ICondensation with background clutter

Period Occasion Description
09/2012 - now Research Scientist within TU9/Sinogerman network project involved sub-projects: Car2x and ECU
September 2013: School of Mobile Information Engineering, Sun Yat-sen University Zhuhai, China. Teaching one week lab-course on programming an electric car using an Altera FPGA
December 2012: 2 Weeks visit at Institute of Microelectronics, Tsinghua University, Beijing Kick-Starting cooperation in GP-GPU computation
03/2009 - 09/2012 IGSSE Scholarship for Ph.D. research in Image Aided Flight Control Research in visual processing for quadcopter
03/2011 - 06/2011: Internship at Willow Garage tight integration of intertial measurements with visual SLAM algorithms
10/2003 - 02/2009 Technische Universität München, Diploma in Informatics passed with distinction

Computer Vision Tools Library (CVT)

CVT is an open-source library for various useful algorithms in the area of computer vision. It mainly served as a research platform for my colleague Philipp Heise and myself. The implementations are tailored to our research needs.

The library is available under MIT license and on github: CVT@github

edvo ROS package

The edvo package is the associated source code for the IROS paper on Efficient Direct Visual Odometry. A more detailed description can be found on the corresponding github page: edvo@github

imu_filter ROS package

The ROS package, developed during my internship at Willow-Garage imu_filter ROS page