ibclogo XVI International Botanical Congess


Abstract Number: 2292
Session = 8.7.7


IMAGE ANALYSIS AS A TOOL IN AIRBORNE POLLEN IDENTIFICATION


Mervi Hjelmroos-Koski,University of Arizona


Quantitative analysis of digital microscopic images of pollen grains has been used to develop a semi-automated identification system for accurate classification of airborne pollen. Results based on the measurements of 16220 images of pollen grains from 80 species demonstrated that it was possible to determine an individual pollen grain's species of origin for 70.9% of the imaged pollen and its family for 76.7% of the grains. A total of 890 measurements, including 20 morphological features, 70 perimeter features and 800 texture features were extracted from each pollen. Identification accuracy's for the 80 species varied considerably with some being classified with nearly 100% accuracy and others to below 50%. Pollen grains of Parietaria, Urtica, Juniperus communis, J.scopulorum, J.ashei, Cupressus sempervirens, C.glabra and Taxus baccata were among the species with the highest identification accuracy.


HTML-Version made 7. July 1999 by Kurt Stüber