Convolutional neural networks applied to the interpretation of lineaments in aeromagnetic data

Parameter estimation in aeromagnetics is an important tool for geologic interpretation. Due to aeromagnetic data being highly prevalent around the world, it can often be used assist understanding geology of area as a whole or locating potential areas further investigation mineral exploration. Methods that automatically provide information such location and depth source anomalies are useful interpretation, particularly where large number exist. Unfortunately, many current methods rely on high-order derivatives therefore susceptible noise data. Convolutional neural networks (CNNs) subset machine-learning well-suited image processing tasks, they have been shown effective at interpreting other geophysical data, seismic sections. Following several similar successful approaches, we developed CNN methodology estimating lineament-type maps. To train model, synthetic modeler vary relevant physical parameters, representative set approximately 1.4 million images. These were then training classification CNNs, with each class representing small range values. We first applied model series difficult sets varying amounts noise, comparing results against tilt-depth method. from northeastern Ontario, Canada, contained dike known was correctly estimated. This method robust easily new using trained which has made publicly available.