Application of wavelet transform in the classification of pollen grains

Diogo Soares da Silva, Lia Nara, Balta Quinta, Ariadne Barbosa Gonçalves, Hemerson Pistori, Marcelo Rafael Borth

African Journal of Agricultural Research · 2014 · 13 citations · 12 references

DOIFull text

Open access

Concepts

Abstract

This paper presents a proposal for the automatic classification of pollen grains, to assist in the analysis of the botanical origin of honey samples. For this task, techniques based on computer vision and machine learning were used. At first, the pollen grain image was segmented using a technique based on watershed. Then, a wavelet transform technique was used to extract texture features from the segmented image. Finally, a supervised machine learning technique was used to classify the pollen grain regarding its floral species. Other attributes based on shape, texture and color were taken for comparison with the proposed method. The technique has been assessed using an image dataset from 7 different pollen species and a 79% F-Score has been achieved.

References

12