Deep Convolutional Auto-Encoder for Automated Processing of Airborne Time-Domain Electromagnetic Data

TY - CONF

T1 - Deep Convolutional Auto-Encoder for Automated Processing of Airborne Time-Domain Electromagnetic Data

AU - Asif, Muhammad Rizwan

AU - Larsen, Jakob Juul

AU - Auken, Esben

AU - Christiansen, Anders Vest

N1 - Conference code: 3

PY - 2022

Y1 - 2022

N2 - Modern airborne electromagnetic surveys produce large datasets that may contain thousands of line kilometers of data. In urbanized areas, parts of these data are often disturbed due to couplings to infrastructure and must be culled before inversion for reliable geological interpretation. As of today, the process of identifying couplings is generally a manual task, which require specialists to examine and process the data subjectively. These subjective workflows are difficult to reproduce and relatively time-consuming. To eliminate the complex data processing workflows, we propose an algorithm based on a deep convolutional auto-encoder to identify coupling in the data in an automated manner. The autoencoder is configured to encode the TEM data in a latent space. The encoded data is trained to decode the encodings to reconstruct the input data. If clean data are observed by the auto-encoder, the reconstructed output will have low error to the input. However, when dealing with couplings, the reconstruction error is elevated, indicating a non-geologic anomaly. We show that the proposed approach displays high quality data processing within a fraction of a second, and is a significant step towards a complete automated processing and inversion workflow, greatly reducing costs and increase reliability of airborne datasets.

AB - Modern airborne electromagnetic surveys produce large datasets that may contain thousands of line kilometers of data. In urbanized areas, parts of these data are often disturbed due to couplings to infrastructure and must be culled before inversion for reliable geological interpretation. As of today, the process of identifying couplings is generally a manual task, which require specialists to examine and process the data subjectively. These subjective workflows are difficult to reproduce and relatively time-consuming. To eliminate the complex data processing workflows, we propose an algorithm based on a deep convolutional auto-encoder to identify coupling in the data in an automated manner. The autoencoder is configured to encode the TEM data in a latent space. The encoded data is trained to decode the encodings to reconstruct the input data. If clean data are observed by the auto-encoder, the reconstructed output will have low error to the input. However, when dealing with couplings, the reconstruction error is elevated, indicating a non-geologic anomaly. We show that the proposed approach displays high quality data processing within a fraction of a second, and is a significant step towards a complete automated processing and inversion workflow, greatly reducing costs and increase reliability of airborne datasets.

UR - http://www.scopus.com/inward/record.url?scp=85161473770&partnerID=8YFLogxK

U2 - 10.3997/2214-4609.202220038

DO - 10.3997/2214-4609.202220038

M3 - Paper

SP - 1

EP - 5

T2 - NSG2022 3rd Conference on Airborne, Drone and Robotic Geophysics

Y2 - 18 September 2022 through 22 September 2022

ER -