Development of Jordan Elman Neural Network for Classification of Pollen Grains Using Histogram based Features

Abstract

A new classification algorithm is proposed for pollen grains, which play an important role in classification of plants from pollen SEM images.The traditional method of pollen classification is tedious and needs experts from the field of palynology. As pollen grain is a complex pattern, its recognition and classification is a challenging problem and hence neural network and computational intelligence approach may be suitable.In this context, neural network and computational approach might prove more suitable. Jordan Elman Neural Network based classifier is found to be optimal with regard to average classification accuracy on CV dataset. A novel scheme of feature extraction comprising of image histogram coefficients has been suggested. The performance of JE NN has been compared with that of MLP NN and SVM. It is shown that the proposed strategy of classification could provide an efficient alternative to the prevalent method of plant taxonomy.