Abstract: This paper describes the problem of automated pollen grains image recognition using images from microscope. This problem is relevant because it allows to automate a complex process of pollen grains classification and to determine the beginning of pollen dispersion which cause an the allergic responses. The main recognition methods are Hamming network Korotkiy, 1992 and structural approach Fu, 1977. The paper includes Hamming network advantages over Hopfield network Ossowski, 2000. The steps of preprocessing (noise filtering, image binarization, segmentation) use OpenCV Bradsky et al, 2008 functions and the feature point method Bay et al, 2008. The paper describes both preprocessing algorithms and main recognition methods. The experiments results showed a relative efficiency of these methods. The conclusions about methods productivity based on errors of type I and II. The paper includes alternative recognition methods which are planning to use in the follow up research.

Keywords: image recognition, OpenCV, Hamming network, feature points method, pollen-grains, structural pattern recognition.

ACM Classification Keywords: I.5.1 Pattern Recognition Model - Neural nets, Structural, I.5.4 Pattern Recognition Applications - Computer vision

Link:

POLLEN GRAINS RECOGNITION USING STRUCTURAL APPROACH AND NEURAL NETWORKS

Natalia Khanzhina, Elena Zamyatina

http://www.foibg.com/ijima/vol04/ijima04-03-p03.pdf