A new approach for border detection of the Dumluca (Turkey) iron ore area: Wavelet cellular neural networks

Abstract

Anomaly analysis is used for various geophysics applications such as determination of geophysical structure's location and border detections. Besides the classical geophysical techniques, artificial intelligence based image processing algorithms have been found attractive for geophysical anomaly analysis. Recently, cellular neural networks (CNN) have been applied to geophysical data and satisfactory results are reported. CNN provides fast and parallel computational capability for geophysical image processing applications due to its filtering structure. The behavior of CNN is defined by two template matrices that are adjusted by a properly supervised learning algorithm. After training stage for geophysical data, Bouguer anomaly maps can be processed and analyzed sequentially. In this paper, CNN learning and processing capability have been improved, combining Wavelet functions and backpropagation learning algorithms. The new architecture is denoted as Wavelet-Cellular Neural networks (Wave-CNN) and it is employed to analyze Bouguer anomaly maps which are important to extract useful information in geophysics. At first, Wave-CNN performance is tested on synthetic geophysical data, which are created by a computer environment. Then, Bouguer anomaly maps of the Dumluca iron ore field have been analyzed and results are reported in comparison to real drilling results.