Magnetic anomaly separation usıng Genetic Cellular Neural Networks

Magnetic anomaly separation usıng Genetic Cellular Neural Networks

Osman N. Ucan¹, Erdem Bilgili² and A. Muhittin Albora³

1 Istanbul University, Engineering Faculty, Electrical & Electronics Department, 34850, Avcılar, İstanbul, Turkey;

E-mail: uosman@istanbul.edu.tr

^{2 TUBITAK Marmara Research Center, P.O:21 41470 Gebze, Kocaeli, Turkey.}

3 Istanbul University, Engineering Faculty, Geophysical Department, 34850, Avcılar, İstanbul, Turkey.

E-mail: muhittin@istanbul.edu.tr

(Received 28 May 2001, accepted 29 October 2001)

Abstract: In this paper, a contamporary stochastic image processing novel, Genetic Cellular Neural Networks (GCNN) is applied the first time in geophysics. The new approach has been applied to gravity anomaly separation problem. The advantages of CNN method are that it introduces little distortion to the shape of the original image by using neighbourhood locations and stochastic properties of 2-D images and that it is not effected significantly by factors such as the overlap power spectra of regional and residual fields. Genetic algorithm is a statistical optimisation technique using a natural selection. In this paper, coefficients of CNN templates A, B and I are trained using genetic algorithm for geophysical data. Here the proposed method is tested using a synthetic examples and satisfactory results have been found.

Keywords: Genetic Cellular Neural Network (GCNN), Magnetic Anomaly