基于差分盒维数的井下电视图像裂缝分割识别方法研究
Research on Crack Segmentation and Identification Method for Downhole TV Images Based on Differential Box Dimension Algorithm
田成富 1耿德祥 1曹建伟 1王明明2
作者信息
- 1. 湖北省地质局 地球物理勘探大队,湖北武汉 430056;湖北省神龙地质工程勘察院有限公司,湖北武汉 430056
- 2. 宿州学院资源与土木工程学院,安徽宿州 234000
- 折叠
摘要
利用井下电视图像对岩心完整性进行识别、分割及裂缝宽度提取是解释岩性、裂隙的关键.目前,井下电视成像结果主要是通过人工方法进行识别解释,工作量大且易受主观因素影响.基于此,针对井下电视图像分割识别,提出从分形理论出发,采用差分盒维数(Differential Box-counting,DBC)算法来提取井下电视图像的岩性变化,以及裂缝、裂隙等岩层纹理特征,进一步通过基于最佳聚类数选择的K-均值聚类算法对井下电视图像进行纹理分割,实现对井下电视图像的岩层自动化分区.结果表明:结合Canny边缘检测算法提取裂缝宽度信息,裂缝识别准确率达到94.2％,实现了岩性的准确分层、裂缝的精细识别.与2D Log-Ga-bor算法相比,差分盒维数算法对井下电视图像分割效果好、速度快.
Abstract
The key to interpreting lithology and fractures is to identify,segment,and extract crack width from downhole TV images of rock core integrity.At present,the imaging re-sults of downhole television are mainly identified and interpreted through manual methods,which requires a large workload and is easily influenced by subjective factors.This paper fo-cuses on the segmentation and recognition of downhole television images.Starting from frac-tal theory,this paper proposes to extract the lithological changes by the differential box counting(DBC)algorithm and rock texture features such as cracks and fissures from down-hole television images.Then,the K-means clustering algorithm based on the optimal num-ber of clusters is used to perform texture segmentation on downhole television images,achie-ving automatic rock layer partitioning of downhole television images.The results show that by combining the Canny edge detection algorithm to extract crack width information,the ac-curacy of crack recognition reaches 94.2％,achieving accurate layering of lithology and fine identification of cracks.Compared with the 2D Log-Gabor algorithm,the differential box di-mension algorithm has better segmentation performance and faster speed for downhole TV images.
关键词
差分盒维数/井下电视/图像裂缝分割识别Key words
differential box dimension/downhole television/image crack segmentation rec-ognition引用本文复制引用
基金项目
安徽省高校自然科学研究项目(2023AH052232)
宿州学院博士后科研启动基金(2022BSH002)
综合地球物理研究所校级平台项目(2021XJPT54)
出版年
2024
NETL
NSTL
万方数据