Imaging system automates pollen measurements
A new type of measuring station based on machine vision and light optical microscopy automatically determines tree pollen count and improves the forecast.
For the 2009 spring allergy season, the German weather service has ordered 15 of these new measuring stations developed by researchers at the Fraunhofer Institute for Applied Information Technology FIT and for Toxicology and Experimental Medicine ITEM, working in collaboration with Helmut Hund GmbH.
A new type of measuring station based on machine vision and light optical microscopy automatically determines tree pollen count and improves the forecast. For the 2009 spring allergy season, the German weather service has ordered 15 of these new measuring stations developed by researchers at the Fraunhofer Institute for Applied Information Technology FIT (Sankt Augustin, Germany; www.fit.fraunhofer.de) and for Toxicology and Experimental Medicine ITEM (Hannover, Germany; www.item.fraunhofer.de), working in collaboration with Helmut Hund GmbH (Wetzlar, Germany; www.hund.de).
Current forecasting techniques are not always reliable. They are based on the weather and the amount of pollen in the air. The problem is that limited data on current pollen levels are available, as these are difficult and time-consuming to obtain.
Typically, ambient air flows onto a piece of adhesive tape and the pollen sticks there. Laboratory workers examine the trapped pollen under a optical microscope and count the different grains. This is a tedious procedure and is only carried out at selected locations. A truly reliable forecast would require a closer-knit network of measuring stations.
The innovative feature is the analysis method: The stations determine the pollen composition fully automatically and transmit the data to the weather service. "To do this, the stations, which are housed in a large container, ingest a controlled amount of air. The pollen grains contained in this air are cleansed of any impurities and deposited on a carrier," says Thomas Berlage, director of Life Science Informatics at FIT.
The pollen-grain carrier--a thin sheet of glass--is covered with a layer of gel. The pollen grains sink into this gel. A light optical microscope automatically takes pictures of the pollen. However, there is a challenge: In these 2-D images, the primarily spherical pollen grains--regardless of whether they come from birch, hazel, or alder trees--are only displayed as circles. When viewed in three dimensions, however, the different types of pollen exhibit differences such as furrows.
"To overcome this difficulty, the microscope examines 70 different layers by automatically readjusting the focus 70 times," explains Berlage. In some views the highest point of a pollen is in focus, in others the center. For each level, the system calculates the points that are most clearly pictured. It then combines all these points to form a 2-D image that contains the 3-D information--the image shows the "flattened" top half of the pollen.
If a pollen grain has a furrow at this point, it can be seen on the image. From this information, the system calculates certain mathematical features, compares these with a database, and determines the type of pollen. The results are available within one or two hours and are transmitted to the weather service via a network connection.