基于区域增长和三维匹配的岩心油水分布统计方法研究

发布时间：2018-12-14 16:00

【摘要】：随着计算机硬件和软件的高速发展,计算机的相关技术广泛的应用在了各行各业,这些技术和相关行业的有效结合,可以大大降低工作量和提高工作效率,作为国家命脉的石油行业必须充分利用计算机的相关技术。石油储存在岩心中微小的孔隙里,大部分的油和水混合在一起,不易开采,通常采用注水法可以使油聚集在一起,提高石油采收率。如何能使油、水分离,统计油和水的分布状况,分析油和水的关系及运动变化规律,对于提高开采效率至关重要。因此,岩心中油水数目统计和油水形状统计在石油勘探作业中起着相当重要的作用,进而成为石油行业的一个研究热点。本论文从油和水的数目和体积及形状两种统计角度,分别提出了各自的算法。一是岩心中油和水的数目和体积的统计。油团或者水团,可以看成是一个连通域。将单独的一个连通域予以一个标记,称之为连通域标记,标记的个数就是连通域的数目,即为油团或水团的数目。实验采用区域增长法进行标记,即从一个种子像素点出发进行跟踪,直到处理完所有的像素点为止。通过和传统的区域统计算法进行比较,区域增长法的统计效果优于区域统计法。二是岩心中油和水的形状的统计。本论文采用三维形状匹配算法来完成实验,形状匹配算法的基石是确定形状表示,特征描述子的选取是确定形状表示的基础。好的特征描述子,其形状匹配精度比较高,因此特征描述子是本文研究非常重要的一个环节。本文预先定义油团和水团的几种三维形状:圆柱体、球体、长方体、薄膜体等等,如果有某些和上述三维形状无法进行匹配的特殊的形体,另行特殊定义。本论文中选取梯度方向直方图作为实验的特征描述子,然后计算这些形状的梯度方向直方图,通过求取三维数据中每个像素的梯度方向,计算各个油团和水团的梯度方向直方图,通过进行比较,从而可以统计油团和水团的形状数目。通过实验得知,基于梯度方向直方图的形状匹配算法在精确度和速度方面优于其它匹配算法,但是在匹配的精确度方面还有很大的改进空间。
[Abstract]:With the rapid development of computer hardware and software, computer related technology has been widely used in various industries. The effective combination of these technologies and related industries can greatly reduce the workload and improve work efficiency. As the lifeblood of the country, the oil industry must make full use of computer technology. Oil is stored in tiny pores in the core, and most of the oil and water are mixed together, so it is difficult to exploit. Usually, water injection method can make oil gather together and improve oil recovery. How to separate oil and water, to calculate the distribution of oil and water, to analyze the relationship between oil and water and to analyze the law of motion change are very important to improve the recovery efficiency. Therefore, the statistics of the number of oil and water and the shape of oil and water in the core play an important role in oil exploration and become a research hotspot in petroleum industry. In this paper, the number, volume and shape of oil and water are analyzed. The first is the number and volume of oil and water in the core. A mass of oil or water can be regarded as a connected domain. A single connected domain is given a label, which is called a connected domain label. The number of labels is the number of connected domains, that is, the number of oil or water masses. The method of region growth is used to mark, that is to start from a seed pixel point and track it until all the pixels are processed. Compared with the traditional regional statistical algorithm, the statistical effect of the regional growth method is better than that of the regional statistics method. The second is the statistics of the shape of oil and water in the core. In this paper, 3D shape matching algorithm is used to complete the experiment. The cornerstone of shape matching algorithm is to determine shape representation, and the selection of feature descriptor is the basis of shape representation. The shape matching accuracy of a good feature descriptor is high, so the feature descriptor is a very important part of the research in this paper. In this paper, several kinds of three-dimensional shapes of oil and water masses are predefined: cylinders, spheres, cuboids, film bodies and so on. If there are some special shapes which cannot be matched with the above three dimensional shapes, another special definition is given. In this paper, gradient direction histogram is selected as the feature descriptor of the experiment, then the gradient direction histogram of these shapes is calculated, and the gradient direction of each pixel in the 3D data is obtained. The gradient histogram of each oil mass and water mass is calculated, and the shape number of oil mass and water mass can be calculated by comparison. The experimental results show that the shape matching algorithm based on gradient direction histogram is superior to other matching algorithms in accuracy and speed, but there is still much room for improvement in the accuracy of matching.
【学位授予单位】：中国石油大学(华东)
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE31;TP391.41

【参考文献】