Application of an Artificial Neural Network for Airborne Magnetic Data Discrimination

Artificial neural networks (ANN) are computational constructs that attempt to mimic the workings of the human brain with respect to the brain’s ability to detect patterns. ANNs can be trained to determine which class an object belongs to based on selected inputs. the objective of this work was to develop and apply an artificial neural network to discriminate ordnance from non-ordnance based on input derived from airborne magnetic data. A training set was constructed for the ANN from existing data acquired over known items: ordnance, exploded fragments, and non-ordnance. the goal was to be able to predict whether the source of a magnetic anomaly was produced by ordnance. the output classes fell into one of two categories. Either the item of interest was unexploded ordnance (UXO) or it was not unexploded ordnance (non-UXO). This latter category included geology, fragments from exploded ordnance, and non-ordnance items. the low-altitude magnetic data were inverted for source parameters (depth, magnetic moment, and orientation) to generate various parameters that served as input to the ANN. using this training data set, weighting coefficients were computed. the reliability of the ANN was verified and validated using a subset of data with known solutions that were not included in the original training set, and the ANN successfully reduced the number of false positives. in the original assessment of this dataset the analytic signal was to prioritize the anomalies and resulted in a ratio of false positives to true items of 42:75 at 50% target detection and 74:150 at 100% detection. using the ANN on this same dataset the ratio of false positives to true items was improved 5:75 at 50% detection and 68:150 at 100% detection. the use of this ANN will improve the efficiency of the airborne data through enhanced target picking and reduced sampling and excavation costs.