Full text loading...
-
Detection of Capacitive Couplings in Ground-Based TEM Data with a 1D Convolutional Neural Network
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 25th European Meeting of Environmental and Engineering Geophysics, Sep 2019, Volume 2019, p.1 - 5
Abstract
Transient electromagnetics (TEM) is a well-established method for imaging of sub-surface resistivity structures. The method is often used from a moving platform providing densely sampled measurements of the sub-surface structures in 2D or 3D. When measurements are performed close to man-made conductors e.g. buried cables and pipes, the data will be contaminated by coupling artefacts, which must be culled from data sets prior to inversion. This can be a labor-intensive task for large data sets and several methods for automation of this process has been investigated. Here we explore the use of a convolutional neural network (CNN) for detection of couplings in data from a towed, ground-based TEM system. The CNN consists of two 1D convolutional layers followed by three fully connected layers. The proposed method is evaluated by comparing with a neural network based approach that have previously been used to detect couplings in airborne TEM data and is found to have better performance. The performance is presently limited by the size of the training data set.