01 SepPhD DissertationSkeletonization of Data for Seismic Inversion, Seismic Imaging, and Machine Learning
Skeletonization of Data for Seismic Inversion, Seismic Imaging, and Machine Learning
  • Shihang Feng ErsE PhD Student, Supervised by Professor Gerard Schuster
  • Sunday, September 01, 2019
  • 04:00 PM - 05:00 PM
  • Al Khawarizmi Building , Bldg 1, Level 3, Room 3119
2019-09-01T16:002019-09-01T17:00Asia/RiyadhSkeletonization of Data for Seismic Inversion, Seismic Imaging, and Machine LearningAl Khawarizmi Building , Bldg 1, Level 3, Room 3119Karema AlaseefKarema.alaseef@kaust.edu.sa

‚Äč

Abstract: This thesis develops skeletonization methods for seismic inversion, seismic imaging, and machine learning to improve both their computational efficiency and accuracy. To obtain a good starting model for anisotropic full waveform inversion (FWI), the simultaneous inversion of anisotropic parameters vp0 and epsilon are initially performed using the wave-equation traveltime inversion (WT) method. Then a transmission+reflection wave-equation traveltime and waveform inversion (WTW) method is presented for a vertical transverse isotropic (VTI) medium where both traveltimes and waveforms are inverted for the velocity model.

The conventional FWI is sensitive to the amplitude mismatch between the recorded and predicted data. To mitigate this problem, multiscale phase inversion (MPI) is presented where the magnitude spectra of the predicted data are replaced by those of the observed data. Moreover, the data are integrated N times in the time domain to boost the low-frequency components. In this case, the skeletonized data are traces with the substituted magnitude spectra so that only the recorded phase data need to be inverted. 

I have developed a velocity-independent workflow for reconstructing a high-quality zero-offset reflection section from prestack data with a deblurring filter. This workflow constructs a migration image volume by prestack time migration using a series of constant-velocity models. A deblurring filter for each constant-velocity model is applied to each time-migration image to get a deblurred image volume. Ithis case the Hessian inverse is approximated by its skeletonized representation, also known as the deblurring operator. To preserve all events in the image volume, each deblurred image panel is demigrated and then summed over the velocity axis.

A fundamental step in aerial image georeferencing consists of determining the location of GPS control markers on the ground. The GPS marker has a unique hourglass shape and its color is dark. To take advantage of these features, superpixels are used as the skeletonized representations of the targets. Then a superpixel-based classification method is applied to the aerial images. The results show that this method quickly extracting the locations of GPS markers from aerial photographs.

 

 


MORE INFORMATION

  • Karema Alaseef
  • Karema.alaseef@kaust.edu.sa

LOCATION

Top