Projects


Our mission is the fusion of robotic technology and optical sensing techniques to create novel tools for our partners. The majority of our projects is conducted in tight collaboration with partners from industry.

Contact: ruether(at)icg.tugraz.at



Ongoing

BioChipFeeding - Wood chip feeding technology of the future for small-scale biomass boilers

The objective of BioChipFeeding is to develop a new wood chip feeding system of the future for small-scale heating plants. A feeding system able to create fuel blends combined with the ability to perform automatic fuel quality screening enables a rather constant quality feed increasing the plant efficency while simultaneously lowering emissions. Technology to automatically detect and remove foreign matter and overlarge objects from the fuel improves the reliability of the system and thereby reduces unexpected plant shutdowns.

Automatic Camera Calibration - Toolbox for the automatic calibration of cameras and camera systems

The geometric camera calibration serves the purpose of determining the intrinsic parameters of a camera and extrinsic parameters of a camera system. We provide a user-friendly and fully automatic calibration pipeline for geometric camera calibration in conjunction with a circular pattern target and automatic feature detection to reduce the user interaction to a minimum.

PROCAM - Active Illumination and Metrology.

Optical sensing on reflective, transparent or untextured surfaces is difficult. Active illumination and computational photography may help in many aspects of industrial computer vision and 3D reconstruction.

IPOS - Object Recognition and Precise Geo-Localization Below the Train.

Goal of the project is the identification of objects beneath a train running at normal cruise speed. Using a multi-sensor setup, the region between rails is scanned in 3D while running at 80km/h. Specific objects are automatically recognized and geo-referenced by means of GNSS and IMU information.

SENS - Object Recognition and Precise Geo-Localization in Rail Infrastructure

Goal of the project is the image-based identification of objects from a train running at normal cruise speed. Using a multi-view camera rig, the close-by area is searched for specific objects of interest. Once found, the exact geo-location is provided by stereo-triangulation and integration of GNSS and IMU data.

Variational Depth Superresolution using Example-Based Edge Representations

Superresolution of single depth image by combining recent advances in example based upsampling with variational superresolution based on an approximated blur kernel.

AIRPLAN - Automated Spatial Analysis from Aerial Image Data for Safeguarding Mass Events

Goal of the project is the spatial validation and securing of events in public space. We allow the validation and simulation of relevant spatial security and safety aspects on the basis of aerial imaging information. A special focus lies on visibility estimation and observer point optimization in large 3D datasets. This enables, for the first time, a large-scale and resource efficient planning before the event to reduce the probability of a crisis, and quick simulation and reaction in case of emergency.

Conditioned Regression Models for Non-Blind Single Image Super-Resolution

Single Image Super-Resolution that does not rely on a fixed blur kernel (e.g. bicubic) during inference.

Dense Variational Scene Flow from RGB-D Data

Dense variational Scene Flow estimation based on novel data terms to increase robustness to illumination changes, acqusition noise or low measurable texture.

Image Guided Depth Map Upsampling

Upsampling of low resolution depth acquisitions in a convex optimization problem using a high resolution intensity image for anisotropic higher order regularization.

ToFusion - Robust Data Fusion for Reliable Time-of-Flight Imaging.

Combination of robust and recent computer vision methods like energy optimization and machine learning with lowlevel signals from the ToF sensor itself, different sensor modalities and time-sequential sensing.

MedEyeTrack - Fully Automated EyeTracking System for Medical Eyetumor Treatment

Eye tracking system for diagnostical imaging and non-invasive radiation therapy of eye tumors.




Past

NARKISSOS - Virtual Dressing Room
Our vision is to create the next generation dressing room. A 3D model of the customer body is automatically generated, animated and dressed in virtual clothes. Finally, an augmented reality mirror-image is shown back to the user on a large display. This will give the user the impression that he or she actually wears the garment shown on the screen.
SILHOUETTE - Object Boundaries in Manipulation and Classification
In this project we exploit object silhouettes for pose estimation, reconstruction and classification of 3D objects.
QUALITY - Microscopy Image Segmentation
Morphometric enumeration of the hallmark features of steatosis and steatohepatitis in murine liver tissues. The problem of robust segmentation of specific structures in microscopy images is addressed.
QUALITY - Weld Analysis in Harsh Environments
Quality analysis of industrial welding is a challenging field for computer vision. The development of a special camera system by an industrial partner provides the opportunity to acquire images of weld seams in realtime, during welding. We address the problems of robust weld seam tracking, mapping and automated defect detection.
MICROSENSE - Adaptive 3D Sensing in Industrial Micro Inspection
Recent projecting technologies allow the use of several small projecting units combined with cameras within a single sensor. In micro metrology, a variety of quality characteristics as 3D geometry, surface inspection or even measurements of transparent objects are investigated. Research on lighting optimization is done to cope with complex acquisition situations and to facilitate automated user guidance.
MVM2 - SLAM - Simultaneous Localization and Mapping
SLAM addresses the problem to build a map of an unknown environment and simultaneously estimate the robots position. In this project we use different types of sensors (stereo-cameras, laser rangefindes, sonar sensor or Time-of-Flight cameras) to solve the SLAM problem through probabilistic or incremental Structure from Motion techniques.
METROLOGY - Rigid Body Reconstruction for Motion Analysis of Giant Honeybees using Stereo Vision
We reconstruct and analyze the social waving of giant honeybees. Project in collaboration with the institute for zoology of the Karl Franzens University Graz. The system is designed to capture the nest of giant honeybees and perform a postprocessing segmentation and tracking of surface bee. The framework returns the positions of the bees in 3D at a temporal resolution of 60 Hz.
MICROSENSE - 3D Image Reconstruction using Active Wavefront Sampling
In this project a 3D surface imaging technique called Active Wavefront Sampling (AWS) is evaluated. In this approach, the optical wavefront traversing a lens is sampled at several off-axis locations and the resulted image motion of each target is measured. This image motion can then be used to calculate the according depth map. Within this project we focus on the estimation of the image motion created by such an AWS system.
MICROSENSE - Adaptive Illumination with Projector-Camera Systems
Optimizing illumination for image acquisition and optical metrology, using novel projector technologies.
MVM2 - 3D Visual SLAM
Visual SLAM is an emerging technology for fast and cheap 3D motion estimation and map building. Within this project we investigate the ability of existing Structure from Motion algorithms to perform large scale indoor SLAM.
METROLOGY - Catadioptric Stereo & Visual Hull
The aim of this project is to compute 3D reconstructions and shapes from mirror images. The advantage of this stereo system is that no synchronization between multiple cameras is necessary. Additionally, geometric relations between camera and mirror(s) can be used in order to calibrate the whole setup. In order to be able to capture mirror images, a prototype will be constructed in the laboratory.
METROLOGY - Pose Estimation of Known 3D Objects using Catadioptric Stereo and Planar Mirrors
Industrial automated manufacturing of objects or components often requires knowledge of their orientation or pose. In this project a setup is presented which computes the pose of a visible object from a single image over catadioptric stereo relations. The underlying software is also able to reconstruct the visual hull. Object pose estimation is done in realtime, which means approximately 200ms per object including positioning by a robot end effector.
MVM2 - Simultaneous Localization and Mapping (SLAM) for mobile plattforms with recent sensor technologies
Simultaneous Localization and Mapping (SLAM) is still a topic of wide interest in the robotics community. This task is essential for autonomous navigation, path planning and obstacle avoidance. SLAM describes a process of building a map of an unknown environment and computing at the same time the current robot position. The aim of the work is to find a robust method to generate 2D or 3D maps with recent sensor technologies.
QUALITY - Fast Method to Measure Fiber Diameter and Diameter Distribution in Melt Blown Polypropylene Webs
Polypropylene nonwovens produced by melt blow technology are widely used in various fields. The quality of melt blown PP nonwovens is determined by web structure, such as porosity and thickness, as well as fiber properties. Within this project novel software is developed to characterize fiber diameter and fiber diameter distribution from SEM pictures of melt blown PP webs. This is done by fiber centerline detection and adjacent radius estimation for a large amount of fibers visible in images of different magnification. Resulting measurements are finally visualized as histogram, showing the fiber diameter distribution.
METROLOGY - 3D Reconstruction and Geographical Referencing of Lightning Discharges
A computer vision setup is presented for the automated detection and multi-view 3D reconstruction of lightning discharges. The system is designed to operate autonomously in a continuous mode and provide accurate position information of the lightning discharge path and impact point relative to its surrounding area. Extensive tests in a medium scale laboratory environment show that a relative measurement accuracy of 1:200 is reachable with VGA resolution cameras.
METROLOGY - A Low-Cost System for Tracking People in a Labyrinth
After medical treatment of visually handicapped people it is desirable to evaluate the benefit of the treatment for the patient. Especially the capability of the patient to orient himself in a three-dimensional environment, to navigate and recognize obstacles is of interest. For a clinical evaluation under controlled circumstances a labyrinth has been built through which the patient ha to navigate. Obstacles may be randomly placed in the labyrinth. A multi-camera system keeps track of the patients movements and extracts parameters such as position, speed, head rotation etc.
METROLOGY - Human Body Volume Measurement
A multiview system is presented for reconstructing a human body to estimate its total body volume. A generic humanoid model is fitted into the three-dimensional point cloud of the underlying human body, from which the whole body volume is derived from.
METROLOGY - Solution of an Industrial Measurement Task with a Robot Assisted Structured Light Sensor
Typical industrial measurement tasks as planarity inspections, distance/diameter measurements and part identification are flexibly performed with a robot assisted structured light sensor.