Former Research
3D reconstruction of barley plants
An important research topic in agriculture is the identification of stress-resistent or high-yield barley genotypes. For identifying such genotypes various different types are sowed and during their growth period pnenotypical numbers like height, number of leafs, or the flowering time are manually aquired in regular intervals. We are working on automated procedures to simplify these monitoring procedure. In particular we seek to reconstruct the plants in 3D from multiple views acquired from different view points. Once 3D data is available all phenotypical data of interest can easily be extracted.
Cooperations:
Plant Breeding Group, Martin Luther University Halle-Wittenberg, Prof. Dr. Klaus Pillen
Dr. Armagan Elibol, at that time with the Korea Advanced Institute of Science and Technology (KAIST)
Mosaic Images
Mosaic images are used to compactly represent image sequence data acquired with active cameras.
Marvin - Computer Vision on a Mobile Platform
Our mobile robot Marvin mainly participates in lectures we offer at the Master's level. Marvin serves as a mobile platform for investigating computer vision algorithms, e.g. dealing with collision avoidance, indoor navigation or human-machine-interaction in general. But, working with Marvin is not restricted to image analysis and computer vision. Besides the Kinect camera offering 3D data of the environment, the sonar and the integrated microphones also allow for work in other areas of signal processing. Furthermore of course Marvin requires an elaborate control system subsuming its logic and planning modules, which is also continously improved.
Statistical Analysis of Microarray Data
DNA-Microarrays, respectively DNA-Chips, are a high-throughput technique for measuring mRNA expression levels. It is used for testing for different cancer types and for analysis of gene function in cell- and tumour research.
Correction of Overshining Effects in cDNA Microarrays
In cDNA-Microarrays the quantitative evaluation of signals of single spots is often severly influenced by strong signals in their neighborhood, which is called overshining and which needs to be compensated for reliable results.
Proteindocking
Predict position and strength of protein-protein interactions from known 3D structures for the rigid and flexibe case.
Perceptual Grouping
Exploit Gestalt laws for perceptual grouping of contour primitives within a grouping hierarchy using Markov Random Fields. Additionally, region information is integrated and dynamic issues of image sequences are treated.
Amperometric Biosensors
Analysis of amperometric signals from biosensors using Hidden Markov Models to determine the concentration of various analyts.