Document Type |
: |
Thesis |
Document Title |
: |
Combining Multiple Seismic Attribute Using Machine Learning دمج عدة سمات زلزالية باستخدام تعلم الآلة |
Subject |
: |
Faculty of Computing and Information Technology |
Document Language |
: |
Arabic |
Abstract |
: |
Seismic exploration is used to estimate the properties of the Earth's subsurface from reflected seismic waves. Geological experts need to examine seismic data to be able to visualize and detect geologic facies and reservoir properties, such as thickness, fluid type, and structure. Combining seismic data and quantities derived from them, called seismic attributes, enhances the resulting visualized data as seismic attributes provide complementary information. Also, combining seismic data and seismic attributes, referred to as the combining task, saves a lot of time and effort, since the number of seismic attributes volumes used in seismic exploration is growing rapidly. Conventional methods manage to combine a limited number of seismic attributes simultaneously. To overcome this limitation, we propose using image Deep Learning-based fusion techniques to combine seismic data and multiple seismic attributes. Deep Learning approach is preferred as it enhances the performance of fusion techniques by utilizing neural network capabilities in feature extraction. The contribution of this research is evaluating the use of six pretrained image fusion models that have achieved the best results in their respective tasks, in the combining task. This is the first study to make use of fusion techniques, as these techniques have only been used to enhance the resolution and reduce the noise of a single seismic attribute. In addition, this is the first study that utilized pretrained models in the combining task. Several techniques have been used to evaluate the results of running those models on public data sets, such as conducting qualitative questionnaires targeting geological experts, and using image fusion metrics. The experiments showed that image fusion techniques have succeeded in the combining task and its results are not distorted. It turns out that the Image-fusion Framework using CNN (IFCNN) model outperformed all other models in both quantitative and qualitative analysis. As comparative study, IFCNN is compared with the current state-of-the-art technique, called Octree quantization method. IFCNN overcomes the limitation of the Octree quantization method and succeeded in combining nine seismic attributes with better fusion quality. |
Supervisor |
: |
Dr. Mai Fadel |
Thesis Type |
: |
Master Thesis |
Publishing Year |
: |
1443 AH
2022 AD |
Co-Supervisor |
: |
Dr.Amani Jamal |
Added Date |
: |
Sunday, January 8, 2023 |
|
Researchers
أبرار مطلق العتيبي | Alotaibi, Abrar Mutlaq | Researcher | Master | |
|