Research On Identification And Prediction Method Of Core Mud Interlayer Based On Image Analysis

The identification and characterization of core data is of great significance in geological research and oil production and development.As the earliest batch of overseas oil sands blocks developed,the Mc Kay River oil sands block in the Lower Cretaceous Mc Murray(MCMR)Formation reservoir is rich in core data.Due to its special tidal-control estuary deposition system,the reservoir has strong heterogeneity,and a large number of millimeter-level muddy interlayers are developed in the core.It is difficult to meet the research needs only by manually identifying the interlayer,and the reservoir permeability anisotropy Characterization is difficult,and it is urgent to use image recognition algorithms to automatically identify and solve the problem of interpretation and distribution of muddy thin interlayers in the study area.Due to the large number of thin muddy interlayers and extremely unstable lateral extension,it has a great influence on the development of the vapor cavity in the development of oil sands.Therefore,it is of great significance to carry out interlayer distribution prediction on the basis of the identification of thin core muddy interlayers.The identification of core and muddy interlayer utilizes the characteristics of the difference between the muddy interlayer and surrounding rock in the core image,and uses image segmentation technology to filter and classify statistics.In recent years,various image segmentation algorithms have emerged in an endless stream,but various new image segmentation algorithms can only be applied to specific places.As a new image recognition object,the core and muddy thin interlayer need to compare different types of recognition methods to select the best Recognition methods serve mezzanine interpretation and prediction.In this study,OTSU segmentation,particle swarm double threshold segmentation,FCM cluster segmentation,and neural network method were carried out to carry out a comparative study of automatic thin interlayer recognition in the study area.The core recognition results of each sample show that the average accuracy of the particle swarm optimization algorithm reaches 90.29%,and the characteristics of simple algorithm,few parameters,good robustness,high reliability,fast convergence,etc.can better achieve large data volume millimeter-level core photos Identification of thin muddy interlayers.Aiming at the combination of the core image of the Mackay Oil Field in Canada and the computer automatic recognition technology,a complete set of core thin interlayer recognition software system architecture and development has been completed.The software can realize the invalid mark removal,image enhancement and image sharpening of the original core photos After pretreatment,it can identify the typical thin muddy interlayers in the classified core photos and make quantitative statistical analysis.The automatic identification of muddy thin interlayers developed at high frequency in 185 coring wells in the study area was completed.Statistics show that the thickness of the interlayer is mainly distributed between2mm-8mm,the termination frequency of different mudstones is between 0.04-20,and the extension length of the interlayer is between 0.1m-46 m.Then,based on the queuing theory,the frequency of mudstone termination frequency is calculated,based on the termination frequency,the extension range of the interlayer of different thickness is predicted,and the quantitative relationship of the extension length of the mudstone interlayer of different thickness is established to realize the prediction of the three-dimensional distribution of interlayer.Better serve oil field production,and provide references for the efficient development of other oil sand projects at home and abroad.