Improved Classification of Pollen Texture Images Using Svm and Mlp

Humans are interested in the determination of the geographical origin of honeybee pollen due to its nutritional value and therapeutical benefits. This task is currently being developed in a manual way using images from optical microscopy. We have proposed [1, 2] an automatic system for pollen identification, based on its texture classification using a minimum distance classifier. In the present paper, we explore the use of more sophisticated classifiers to improve the classification stage. Specifically, we apply several well-known classifiers, KNN, Support Vector Machine and Multi-Layer Perceptron, in order to increase the classification rate on this problem.