Neural Network Interpretation Of High-Frequency Electromagnetic Ellipticity Data

This paper focuses on the detection of three dimensional (3D) small conductive targets from high-frequency<br>electromagnetic elliptic@ data using neural networks. For environmental investigations it is necessary to provide<br>as much information on the location of shallow buried conductive objects or the electrical properties of possible<br>contaminants. The networks are trained with one-dimensional (ID) forward models to estimate the resistivity and<br>dielectric constant structure of the ground. The input is given by elliptic@ sounding curves from eleven discrete<br>frequencies in binary steps in a range from 32 kHz to 32 MHz. Halfspace and layered earth neural networks will<br>provide reasonable fit to sounding curves even if they are influenced by shallow conductive 3D objects. We show<br>that a detailed inspection of elliptic@ profiles over targets such as a 5 m by 3 m aluminum sheet (depth of 1 m), a<br>5%gallon barrel (depth of 0.63 m), and two metal desk (depth of approximately 1 m) can help to detect these<br>anomalies. Piecewise halfspace neural network are capable of enhancing the anomalies in resistivity depth sections<br>and provide additional information for the detection and possible localization of the object. The visualization of<br>the results is very important since small targets will show up as subtle anomalies. Based on observations of<br>elliptic@ sounding curves and profiles we can train a neural network to classify target responses versus<br>background responses for specific sites, assuming that enough soundings are available to train the neural networks.