图像分割在区分铸体薄片不同组分中的应用

The Application of Image Segmentation in Distinguishing Different Components of the Casting Slice

摘要: 铸体薄片是分析孔喉结构最常用也最直接的方法，但目前多采用描述或半定量方法。为精细评价孔喉结构等特征，实现高效可靠的岩石分类，提出了基于Otsu方法与决策树自动拾取多个阈值和采用人工经验与计算机结合拾取目标像素域两种方法，分割铸体薄片图像提取不同岩石组分。采用吐哈盆地红台构造带的10幅铸体薄片图像对两种方法进行测试，并对比了从分割后的铸体薄片图像中所提取面孔率和铸体薄片实验提供数据。结果表明：两种方法可以有效区分铸体薄片中的不同组分，基于Otsu方法与决策树自动拾取多个阈值的方法可以高效的分割铸体薄片图像组分，但无法自动区分目标；采用人工经验与计算机结合的方法，在半自动化分割图像的基础上可自动提取目标，对目标定量化评价。

Abstract: The casting slice is the most common and direct method to analyze the pore-throat structure; however, most of them are descriptive or semi-quantitative. In order to accurately evaluate the characteristics of pore-throat structure and achieve efficient and reliable rock classification, two methods are proposed. The method of automatically picking up multiple thresholds based on the Otsu algorithm and the decision tree, or using the artificial experience with the computer to pick up the target pixel domain, thereby segmenting the casting slice to extract different rock components. The two methods were tested by using 10 casting slices from Hongtai tectonic belt of Tuha basin. I extracted the areal porosity from the segmented image of the casting slice, and compared it with the slice experimental data. The results show that the two methods can effectively distinguish the different components in the casting slice. The method of automatically picking up multiple thresholds based on the Otsu algorithm and the decision tree can effectively segment the image components in the casting slice, but cannot automatically distinguish the target. Using the method of combining artificial experience with computer, the target can be extracted automatically and evaluated quantitatively on the basis of semi-automatic image segmentation.