Position detection of unexploded ordnance from airborne magnetic anomaly data using 3-D self organized feature map

Abstract

The detection of any buried object depends on the analysis of its emerged magnetic anomaly data. In this paper, the self-organized feature map (SOFM) neural network is used to detect the position of the unexploded ordnance (UXO) from airborne magnetic anomaly data. The SOFM is an unsupervised-competitive neural network that can learn a topology-preserving mapping from a high dimensional input space to a lower dimensional map lattice. The SOFM was trained using two-dimensional theoretical magnetic signatures of an equivalent UXO dipole source in different X, Y, and Z coordinates. The feature map is designed as a three-dimensional SOFM, where each dimension in this feature map is used to detect one coordinate of the UXO position. The experimental results showed that the SOFM could accurately recognize the UXO position from both simulated magnetic anomaly data and actual airborne magnetic field data. The network could also recognize the UXO position correctly when tested with noisy magnetic anomaly data up to +/-15%. It is concluded that the SOFM is an efficient, fast, and accurate technique for position detection of the UXO buried objects