Seismic Fault Interpretation Using Deep Learning-Based Semantic Segmentation Method

Hu, Guang; Hu, Zhengwang; Liu, Jiangping; Cheng, Fei; Peng, Daicheng

doi:10.1109/lgrs.2020.3041301

Cited by 22 publications

(5 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Feng et al [17] used Monte-Carlo sampling in the form of dropout at evaluation time to sample multiple possible network predictions and determine the certainty of the model. Hu et al [18] turn to a different convolutional architecture, which, unlike the UNet, does not down-sample and up-sample the data. Instead it uses dilated convolutions and atrous spacial pyramid pooling to analyse the data at multiple resolutions.…”

Section: B Fault Segmentation From Seismic Volumesmentioning

confidence: 99%

Negation Invariant Representations of 3-D Vectors for Deep Learning Models Applied to Fault Geometry Mapping in 3-D Seismic Reflection Data

Kluvanec

McCaffrey

Phillips

et al. 2023

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

We can represent the orientation of a plane in 3D by its normal vector. However, every plane has two normal vectors that are negatives of each other. We propose four novel representations of vectors in 3D that are negation invariant and can be used by a neural network to predict orientation. Our proposed solution is the first to introduce representations that are negation invariant, continuous and easily parallelisable on the GPU. We evaluate the representations by predicting the orientation of a plane on a toy task, and by applying them to synthetic seismic tomographic data where we predict the presence and orientation of faults for every voxel in the volume. We further make use of the orientation of the faults in a post-processing algorithm on the GPU that separates the faults into segments (i.e. instances) that do not intersect, which allows us to selectively visualise faults in 3D. We demonstrate the utility of the representations by deploying the model on the Laminaria 3D Seismic volume as a case study. We quantitatively compare the model's prediction against human interpretations of slices through the volume as well as existing interpretations in literature. Our analysis shows good agreement (F1 score of 88%) of the model with human interpretation in the shallow levels, where the ambient noise is lower, but this agreement degrades at deeper levels (F1 score of 68%). We explore possible reasons for this degradation.

show abstract

Section: B Fault Segmentation From Seismic Volumesmentioning

confidence: 99%

Negation Invariant Representations of 3-D Vectors for Deep Learning Models Applied to Fault Geometry Mapping in 3-D Seismic Reflection Data

Kluvanec

McCaffrey

Phillips

et al. 2023

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Then, we improved and optimized the network, constructing a deep learning semantic segmentation network that can directly realize the end-to-end recognition of surface cracks in high-resolution UAV images. The VGG network series is a very mature network structure that has shown outstanding results in image classification and object detection tasks [27].…”

Section: Deep Learning Network Modelmentioning

confidence: 99%

Recognition of Earthquake Surface Ruptures Using Deep Learning

Chen

Liu

2022

Applied Sciences

View full text Add to dashboard Cite

Investigating post-earthquake surface ruptures is important for understanding the tectonics of seismogenic faults. The use of unmanned aerial vehicle (UAV) images to identify post-earthquake surface ruptures has the advantages of low cost, fast data acquisition, and high data processing efficiency. With the rapid development of deep learning in recent years, researchers have begun using it for image crack detection. However, due to the complex background and diverse characteristics of the surface ruptures, it remains challenging to quickly train an effective automatic earthquake surface rupture recognition model on a limited number of samples. This study proposes a workflow that applies an image segmentation algorithm based on convolutional neural networks (CNNs) to extract cracks from post-earthquake UAV images. We selected the 16-layer visual geometry group (VGG16) network as the primary network architecture. Then, we improved the VGG16 network and deleted several convolutional layers to reduce computation and memory consumption. Moreover, we added dilated convolution and atrous spatial pyramid pooling (ASPP) to make the network perform well in the surface crack identification of post-earthquake UAV images. We trained the proposed method using the data of the MS 7.4 Maduo earthquake and obtained a model that could automatically identify and draw small and irregular surface ruptures from high-resolution UAV images.

show abstract

“…When trying to identify the possible presence of salt deposits underground using seismic data, it's essential to extract specific geological features that are crucial for detecting salt formations. Traditional seismic fault interpretation methods rely on manual identification and attribute analysis, which are labour-intensive and time-consuming [8]. Relying solely on attributes sensitive to salt boundaries, such as instantaneous amplitude, often leads to inaccuracies in interpreting salt formations.…”

Section: Section 1: Introductionmentioning

confidence: 99%

Automatic Salt Segmentation Using Deep Learning Techniques

Jadhav,

Shah,

Vaghani

et al. 2024

Preprint

View full text Add to dashboard Cite

Detecting salt boundaries in seismic images is critical for subsurface reservoir characterization and oil and gas exploration. The presence of salt in seismic data often indicates the presence of valuable hydrocarbon resources, which could lead to significant oil discoveries. Traditional manual interpretation methods have limitations, prompting the industry to embrace deep learning techniques. Our proposed system extensively evaluates the effectiveness of deep learning models in salt boundary detection. Our approach involves developing a custom residual encoder-decoder model and comparing it against two existing models: Res-UNet and UNet. The advantage of our custom-built residual encoder-decoder model lies in its utilization of transposed convolutions for image segmentation. Unlike regular convolutions that extract features and reduce image size, transposed convolutions expand the image, potentially introducing new information from seismic data. The custom model emerges superior to UNet and Res-UNet models, exhibiting an accuracy of 94.28% and a precision score of 0.86. A series of comparative analysis is drawn with a main focus on transforming the automated salt segmentation process.

show abstract

Seismic Fault Interpretation Using Deep Learning-Based Semantic Segmentation Method

Cited by 22 publications

References 19 publications

Negation Invariant Representations of 3-D Vectors for Deep Learning Models Applied to Fault Geometry Mapping in 3-D Seismic Reflection Data

Negation Invariant Representations of 3-D Vectors for Deep Learning Models Applied to Fault Geometry Mapping in 3-D Seismic Reflection Data

Recognition of Earthquake Surface Ruptures Using Deep Learning

Automatic Salt Segmentation Using Deep Learning Techniques

Contact Info

Product

Resources

About