Study on Key Techniques for Sandstone Slice Image Segmentation
|Course||Communication and Information System|
|Keywords||sandstone slice image image segmentation edge flow scale space level set method support vector machine|
ABSTRACT The sandstone is abroad in our country. Oil and gas contained in the sandstone are the main compositions of raw materials used in energy and chemical engineering. So it is significant to study the structure of the grains and the pores between the grains in the sandstone for proving up and exploiting oil and gas. It is an important method to study the reserve of oil and gas from microstructure by analyzing the swatch of sandstone slice through microscope. The sandstones’ pores and granularities can be analyzed quantificationally through micrograph to get correct basic data of microstructures. The key to analyze sandstone slice image is an image segmentation problem. That is how to extract pores and grains from the image correctly. Few reports are about this topic both at home and abroad. This dissertation tries to solve the problem of how to segment sandstone slice images.Firstly, the concept of edge flow is introduced. This thesis proposes how to make intensity edge flow, texture edge flow and phase edge flow. And it proposes how to integrate these edge flows in L~*a~b~* color space.Secondly, the concept of scale space and its properties including causality, morphological invariance and transformation invariance are introduced. And the way to add edge flow vector fields from different scales together by weight is also proposed.Thirdly, the dissertation studies level set method used for curve evolution. It introduces the numerical computation of level set method including a few commonly used fast computation algorithms and the problems that should be concerned when implementing level set method. And it proposes how to segment images using curve evolution under the control of edge flow.