The Application Of Image Segmentation To The Interpretation Of Magnetotelluric Materials

Magnetotelluric(MT) is a method with well prospect for the detection of the earth deep structure and state of material. It plays a significant role in the earth electromagnetic induction branch of Geophysics. People can obtain the electrical structure of the crust and upper mantle in observation area through further processing and inversion towards the data acquired by MT method. And by analyzing the electrical structure, the geologic structure of observation area will be acquired preliminarily, and then the position of subsurface structure edge, including faults and anomalous bodies.Through the data collection, processing and inversion, the ultimate result is a series resistivity map of target areas. And the judgment of geologic structure in these areas derived from analysis towards these map. Therefore, it is necessary to introduce Digital Image Processing to the interpretation process so that artificial deviation can be overcome to some extent. Image Segmentation is a critical area of DIP, and several mature methods have been developed in other fields so far. Modifying these methods and applying them to resistivity map so to support the deside of edge is the content of this paper.I have searched established methods in Image Segmentation, and have screened out some methods with development prospects according to the property of resistivity map, including Region Splitting and Merging, Region Growing, Morphological Watersheds and Relaxation method. Meanwhile, I have researched auxiliary tools that can be used when implementing these methods, including edge detectors and morphological processing. These methods above have been carried out preliminarily via MATLAB language. Then I established a series of typical subsurface structure models, obtained MT forward response, finally acquired inversion result. Those algorithms were modified according to the property of resistivity map. Finally, different segmentation results through these methods have been compared to the original models.Through the discussion above, thesis put forward a method of region expansion with self-adaption. To some degree, we can acquire electrical edge and support geological interpretation through it.