Wyniki wyszukiwania - BazTech

Ograniczanie wyników

1 Journal of Applied Computer Science

1 2009

Znaleziono wyników: 1

Liczba wyników na stronie

Wyniki wyszukiwania

Sortuj według:

Ogranicz wyniki do:

{GA}-Based Feed-Forward Neural Network For Image Classification: Application For the Grains of Pollen

Khorissi N., Mellit A., Guessoum A., Mesaouer A.

Journal of Applied Computer Science

2009

Vol. 17, nr 2

83-96

Palynological data are used in wide range application, the pattern recognition of pollen grains is very important in the determination of the original floral honey, and the prediction of the allergy, which touches lot of people. Due to the high computing time needed for classification of the pollen grains and complex architecture of the neural networks, the genetic algorithm is used in order to find the optimal architecture of the Multilayer perceptron (number of hidden layers and the number of neurons within each hidden layer). In this paper, a methodology for pollen grain classification based on the using of the MLP optimized by genetic algorithm (GA) called MLP-GA is described. A database of pollen images has been used in this work. Firstly, for each image we have calculated some morphological and geometric features. Subsequently, the MLP-GA network has been used for classification of the image pollen. The best classification performance is achieved by using an experimental data base of grains pollen. The classification rate is 90% which is very promising, and by comparing the hybrid MLP-GA with others ANN architectures, we can note that the MLP-GA is faster (the convergence time is 500 iterations).