26 Novembre – Thesis defense - Moctar Mounirou Arouna Lukman
10 h30 Amphi B (Campus de Talence)
Automatic construction of relative geologic time images from seismic images by energy minimization.
Seismic imaging is very important in oil and gas exploration because it allows to represent the subsurfaces of a prospected region on depths up to several kilometers. The obtained seismic images are analyzed during seismic interpretation by geologists and geophysicists in order to understand the basin evolution and locate traps that contain hydrocarbons. In this context, from the seismic image, identify the deposit sequence of geological layers and order them is a very important task. With advances in computational power, several algorithmic solutions have been proposed in recent years in order to perform automatically this so called chronostratigraphic analysis. One of them is the construction of Relative Geologic Time (RGT) images from seismic images. A Relative Geologic Time image, also called GeoTime image, is an image which associates a relative geologic time to each seismic image sample. This allows to date geological layers from the oldest to the youngest. Unfortunately, GeoTime images construction from seismic images that contain noisy regions or complex geologic settings such as unconformities and faults is complicated.
This work aims at proposing new methods for constructing GeoTime images that respect basic principles of geology, and satisfactorly take seismic unconformities and faults into account. The proposed method is based on the minimization of a regularized energy functional. The latter involves the local seismic orientations and a regularization term that allows to introduce various prior information such as geometric deposit models or geological constraints introduced by a geologist. Compared to the state of the art, one of the advantages of the presented method is to provide supervised approaches by taking into account informations such as horizons and faults picking. Furthermore, in order to improve the robustness of the method, we study the case where the energy functional uses the L1-norm as penalty function instead of the L2-norm commonly used. The minimization of the proposed energy results in optimization problems whose solving require the use of proximal algorithms.
