Sie sind hier: Startseite Daten Lehrstühle Mustererkennung und Bildverarbeitung OMNIBUSS


There are about 10 million people in Germany suffering from pollen allergies and this number is increasing each year. An accurate and prompt information about pollen concentration in the air is indispensable to pollen forecast and air pollution studies. Conventionally, pollens are counted manually through microscopes. This task is demanding and time-consuming, and generally cannot be finished in real time. The quality and reliability of the delivered data also vary considerably with the pollen counters.

The joint project OMNIBUSS resulted from these problems. The main purpose of this project is to research on new methods for fast and automatic monitoring of airborne pollens, spores and other allergy-relevant dust particles. This project runs between 1.7.2003 and 31.6.2006. At the end of this project, a marketable online monitor called MICROBUS ( Microscopical Identification and Computerbased Recognition by Online Biological Unit Sampling) should be realized, which includes the following functionalities:

  • Continuous sampling combined with sample preparation for microscopic evaluation.
  • Separation of dust particles of interest in the sample with autofluorescence, color-measuring, etc.
  • 3-D imaging (tomography) for more information of outer and inner structures of the particles.
  • Automatic identification of the particles using pattern recognition techniques.
  • Preparation and transmission of data to a remote data center for further distribution.


Model of the MICROBUS

As one partner of the project OMNIBUSS, we are responsible for developing robust image recognition algorithms for automatic realtime identification of pollens, spores and other airborne particles for the planned online monitor.

In a former joint project with the German weather Service and the Meteoswiss, we have applied image recognition techniques based on gray scale invariants [Schulz-Mirbach:1995] and support vector machines to pollen identification under ideal conditions (good surface separation, 3D images obtained with a confocal laser scan microscope, homogeneous sample material, etc.). The recognition rate is 95% with 26 different pollen species, and 97.4% when all allergically irrelevant species are considered as one class. In OMNIBUSS, our aim is to improve and adapt the image recognition techniques and integrate them into the online monitor. Some of the main problems are:

  • 3D image acquisition. Instead of the expensive Laser Scanning Microscopes (LSM), Conventional fluorescence microscopes will be used combined with structured illumination or subsequent deconvolution.
  • Object segmentation based on gray-scale invariants.
  • Automatic selection of features. As the online monitor will be employed for different tasks, it would be desirable if the most suitable features could be automatically searched.
  • Improvement of robustness of the algorithms for data with reduced quality. The reduction of quality comes mainly from two sources: real-world samples with deformed and contaminated objects instead of laboratory-prepared "perfect'' samples; conventional microscopes in contrary to high-quality and high-priced LSM.


Benutzerspezifische Werkzeuge