Research On Automatic Recognition Technology Of Crack Image In Outcrop Area Of Strata

Crack is a discontinuous body formed by the loss of binding force or structural deformation between rocks.It can not only store oil and gas,but also provide seepage flow for underground fluids.Therefore,reservoir cracks have always been a key point in geological research.At present,the methods for researching reservoir cracks can be divided into two types: geological identification methods and geophysical identification methods.Among them,the geological identification methods mainly include core observation method,field outcrop observation method,microscopic observation method,and development data analysis method;Geophysical identification methods are mostly based on conventional logging,image logging,and seismic methods.Comprehensive application of these crack identification methods and mathematical methods and computer technology,can study and predict the development characteristics and distribution of cracks.And observing the outcrop areas is the most intuitive way to research the morphological characteristics of cracks.Observations and researches through outcrop areas can provide intuitive,visual,and quantitative results in order to understand the evolution and structural characteristics of stratigraphic structures,which is an important means in the study of stratigraphic structures.In the research of traditional outcrop areas,the identification is mainly based on the human eye,to achieve describe the structure of the outcrop areas.However,in actual work,it is very difficult to describe the structure and connectivity of cracks,which is required patience and meticulous long-term work.And human errors can easily occur,leading to inaccurate descriptions.Therefore,a method that automatically recognizes crack images in outcrop areas of strata is presented.The method is divided into three steps: Firstly,in order to remove the shadows present in outcrop area images and enhance the gray-scale contrast of the cracks,the illumination compensation model and a geometric shadow removal algorithm are proposed.Secondly,the tensor voting algorithm is used to enhance the fracture edge information,and a crack saliency map is constructed through the proximity and continuity of the crack pixels.Thirdly,the minimum spanning tree(MST)is constructed to describe the possible connections of the fracture seeds,and a threshold method is used to trim the unwanted edges in the tree to achieve obtaining the final cracks curve.By applying the methods to a real-world crack image in outcrop areas of strata through MATLAB,the cracks are identified correctly.The optimized parameters for the recognition are:the number of partitions in the geometry removal shadow N is 240,the tensor voting scale ? is taken as 25 and the path length threshold L_p is taken as 25.Automatic recognition of crack images provides a technical means for field geological work.Recognizing automatically the digital image of the outcrop in the field can greatly simplify the work intensity of the geological staff,and reduce inaccurate description of the geological structure due to human factors.Simultaneously,it is beneficial to input the structure into the numerical simulation program after recognizing the crack structure,to do a deep research on the evolution and destruction process of the stratum.