About Us

The main area of the research group is three-dimensional (3D) machine perception. We primarily process camera images, depth sensors, and LiDAR (laser scanner) devices. Our main research topics include:

•         Calibration of various sensors, especially LiDAR devices and digital cameras.    
•         Stereo vision using affine transformations.    
•         Visual odometry with vehicle-mounted sensors.

The application of affine transformations is at the center of our research; this topic originated at ELTE. Our main application area is autonomous vehicles. We have developed a sensor suite at the university, which includes cameras with narrow and wide-angle lenses, LiDAR sensor(s), an IMU (Inertial Measurement Unit), and RTK-GPS (Real-Time Kinematic Global Positioning System). The sensors are time-synchronized. A remotely controlled electric go-kart was also built here.

We also developed a programmable robotic arm aimed at automatically unpacking boxes. For this, our research group created a 2D LiDAR-based measurement software capable of determining the size of the object grasped by the robot. Every year, our group organizes a competition for students, where they have to solve tasks related to autonomous vehicles. Our research is supported by Robert Bosch GmbH.

Research Interests

▶ Stereo and multi-view 3D machine vision

•         3D reconstruction    
•         Structure from Motion / Simultaneous Localization and Mapping (SLAM)    
•         Visual odometry    

▶ Multi-modal calibration

•         Camera – LiDAR    
•         Multi-LiDAR    
•         2D LiDAR    
•         RGB-D (color + depth) sensors    

Research Methodology

We believe that in modern research, theory and practice meet. Therefore, we investigate problems that also occur in real-world industrial applications. We prefer collecting our own test data rather than just using data downloaded from the internet. To this end, we developed (i) the ELTECar, a car equipped with numerous sensors, and (ii) the ELTEKart, a remotely controlled electric go-kart.

Infrastructure

•         ELTECar test car and ELTEKart remotely controlled go-kart    
•         16-channel Velodyne VLP-16 LiDAR    
•         HikVision 2 MPixel color cameras with standard and fisheye lenses    
•         RTK correction GPS    
•         Intel Intellysense depth camera    

Research Stuff

•         Levente Hajder, Senior Research Fellow (Research Group Leader, MTMT)    
•         Lajos Lóczi, Senior Research Fellow    
•         Iván Eichhardt, Senior Research Fellow    
•         Bandó Kovács, Engineer    
•         Tekla Tóth, PhD student    
•         Tamás Tófalvi, PhD student    
•         Máté Cserép, PhD student    

Projects

•         3D reconstruction from rectified stereo images using affine transformations    
•         Machine vision for an unpacking robot    
•         LiDAR-camera calibration using spherical calibration objects    

•         Calibration of multiple 2D LiDAR sensors    
•         Homography estimation for offside detection using the center circle    

Key Publications

• T. Tóth, L. Hajder (2023): A minimal solution for image-based sphere estimation, International Journal of Computer Vision [DOI]
• L. Hajder, D. Baráth (2020): Relative planar motion for vehicle-mounted cameras from a single affine correspondence, IEEE International Conference on Robotics and Automation [DOI]
• D. Baráth, I. Eichhardt, L. Hajder (2019): Optimal multi-view surface normal estimation using affine correspondences, IEEE Transactions on Image Processing [DOI]
• D. Baráth, L. Hajder (2018): Efficient recovery of essential matrix from two affine correspondences, IEEE Transactions on Image Processing [DOI]
• D. Baráth, T. Tóth, L. Hajder (2017): A minimal solution for two-view focal-length estimation using two affine correspondences, IEEE Conference on Computer Vision and Pattern Recognition [DOI]

Contact

Research Group Leader: Hajder Levente – hajder@inf.elte.hu
Contact person: Tófalvi Tamás – z2rciu@inf.elte.hu