Application Research Of Unsupervised Clustering Algorithm In Rock Image

With the improvement of the quality of life and the rapid economic growth,people's demand for natural gas and oil is also growing.As the difficulty of high-intermediate gas reservoirs increases,tight sandstones become the main replacement resources.The gas-bearing strata in the Sugri gas field in the Ordos Basin are mainly in the Upper Paleozoic and Shanxi Formations,both of which are tight sandstones.The characteristics of tight sandstones are quite different from those of sandstones.Therefore,further understanding of the microscopic characteristics and macroscopic characteristics of tight sandstones provides some theoretical data for subsequent drilling,operation and mining of tight sandstone gas reservoirs,which has important practical significance for promoting the development of China's petroleum industry.The main work of this paper is as follows:Firstly,the preliminary research on tight sandstone is carried out,and the research status of tight sandstone gas reservoir is expounded.Secondly,the stratigraphic characteristics,rock diagenesis and main pore types of the Upper Paleozoic are described.Then the unsupervised clustering FCM algorithm and K-means algorithm and the improved algorithm of FCM are introduced.In this paper,the histogram-based FCM algorithm can greatly reduce the running time of the algorithm.Combined with the FCM algorithm of rock image color texture features,the segmentation results can clearly distinguish the rock background and rock pores.Finally,the edge segmentation algorithm,histogram threshold method and histogram improved FCM algorithm and FCM algorithm combined with rock image color texture feature are used to cluster the three types of rock images.The segmentation result explains the segmentation effect and the algorithm operation efficiency from two aspects.The results show that the FCM clustering algorithm based on histogram has the shortest running time and good segmentation result.The FCM algorithm combining color features and texture features has good segmentation results,but the algorithm takes a long time.Both methods can well rock.The background is separated from the pores,which lays the foundation for the lithology analysis of the rock.