Organic Matter Segmentation From Images of Scanning Electron Microscope for Shale Gas Reservoir Discovery

The micro-pore characteristics of shale gas reservoirs are the focus of unconventional oil and gas exploration. The shape, size, and connectivity of pores determine the reservoir performance of shale reservoirs to a large extent. Therefore, shale gas exploitation requires a complete understanding of pores. At present, it is an important method to obtain pore structure parameters of shale based on the threshold segmentation method. However, some problems still need to be solved, such as small analysis areas, poor accuracy, and weak quantitative analysis caused by human factors. To solve these problems, an intelligent segmentation technique based on energy dispersive spectrometer (EDS) analysis data is proposed in this paper to achieve intelligent segmentation of organic matter region in Scanning Electron Microscope (SEM) images. Identifying organic matter can lay a foundation for the intelligent identification of shale pores. In this paper, EDS analysis technology is introduced, EDS analysis images are used as label images, and a data set including 1943 EDS analysis images and corresponding SEM images of shale is established. We use U-Net and nested U-Net for organic matter intelligent prediction. Using an SEM image of a shale gas reservoir, intelligent segmentation of organic matter region based on EDS data learning is realized. The results show that the method has high accuracy, can reduce the intervention of human factors, improve the efficiency of analysis, and make the analysis results more accurate and scientific. At the same time, the degree of automation and generalization ability are higher than traditional forecasting methods.