Pollen classification using RBF networks

In this paper pollen cell classification that plays an important role for many applications is achieved by using Radial Basis Function Networks (RBF). Pollen images highly contain texture information that leads us to extract two different types of texture features for classification. The first type features are; angular second moment, entropy, contrast, inverse moment and inertia of the cooccurrence Matrix (CM) obtained form each image and the second one use nine features obtained by Local Linear Transforms (LLT). RBF networks which are known as having good learning capacity are used for classification. In experimental results Bangor/Aberystwyth Pollen Image Database is used. The best classification performance it is achieved by using CM based features and it is 83%. As far as we know, this performance is better than the previous reported results on this database.