Study On Image Processing Technique In Measurement Of Concrete Cracks And Specimen Deformation In Dynamic Triaxial Test

Crack monitoring of concrete and transformation measurements of earth structures are central subjects in civil engineering research.Due to the complexity of the actual environment,both existing crack recognition and transformation measurement technologies are far from meeting the requirements of current engineering practice.In addition,integrated processing technologies such as crack recognition,classification,and analysis are also in urgent need.Dynamic tri-axial experiments are the necessary means to obtain the mechanical properties of soil under dynamic loading.In such an experiment,a displacement sensor with lower transformation precision and credibility measures the transformation of the soil sample,and the local transformation shape of soil sample cannot be obtained.In this paper,crack recognition of concrete and soil sample transformation measurements of a dynamic tri-axial experiment are utilized as study objects.Digital image measurement technology is adopted and crack recognition,image mosaic,crack classification,transformation,and transformation measurement method of soil dynamics are employed to conduct the study.The study contains the following:(1)Enhancement and recognition of the crack image.This study proposes a S-shape fuzzy affiliation function for the parameterization,and furthermore adopts artificial fish swarm algorithms and the least square method to solve unknown parameters,then introducing these parameters into the S-shape function to achieve enhancement of low gray value crack images.An image enhancement algorithm based on pixel classification is introduced,where the crack image is divided into different areas according to area gray variance and gray values,and different affiliation functions are provided to enhance images in different areas to solve the non-uniform gray distribution of the background image.Via calculation of the perimeter and area of the image of the linear crack,the rule of this linear crack based on perimeter-area is obtained.In combination with the extraction of the crack skeleton line,it achieves recognition of linear cracks in heavily polluted background.The proposed crack image segmentation method combines shape feature with gray-mean feature,adopting and 2-D maximum entropy to construct a pixel classification model.Thus the shortcomings of the image segmentation method are overcome that relied solely on a discrete gray value.(2)Crack image mosaic and morphological classification.A fast images matching algorithm is introduced based on Sift feature,adopting a modified SP-Tree structure to search for feature points with most neighboring matches.It has been verified that the reduction of a part of the feature point matchings could achieve fast matching of image feature points.The gray variance of the image block is used to simply describe the crack image,and to combine the length difference of the crack projection,adopting improved particle swarm optimization to build the classification model based on an extreme learning machine(ELM)led to an improvement of the classification accuracy of the whole crack sets and small share of cracks.A polygon fitting method on the edge of the crack is proposed,combined with the Delaunay algorithm,thus generating a finite element mesh to achieve No-slot Joint with CAE analysis software.(3)Modification of digital image correlation algorithm and interpolation model of displacement field.Correlation searching algorithms are introduced based on an improved particle swarm and improved artificial bee colony algorithm,thus increasing the efficiency of relevant digital image search.In addition,radial basis function neural network(RBFNN)is adopted to achieve an entire displacement field interpolation calculation,threshold value Door,and fuzzy C mean clustering algorithm to determine the node number of the center of the hidden layers.The improved firefly optimization algorithm is used to determine the link weight between hidden layers and output layers,followed by the construction of a displacement interpolation model and verification of the effectiveness of this method via calculation of displacement field interpolation of the concrete crack area.(4)Realization of soil sample transformation for digital image measurement technology of dynamic tri-axial experiment.A high-frequency digital camera and high-speed disk storage array is introduced here,and the soil sample transformation digital image measurement technology of geotechnical static tri-axial experiment is applied to the dynamic tri-axial experiment.The traditional dynamic tri-axial pressure chamber is reorganized and a 360-degree surface image of the sample was obtained through a set of plane mirrors and a camera.This achieves synchronization of the collection of displacement sensor and photographing of the camera.A real-time transformation measurement of the whole surface of the soil sample under dynamic load was achieved via corner point recognition technology.Dynamic transformation test and dynamic failure test are carried out.Comparing the strain data measured via image measurement system with the data measured via displacement sensor verifies the advantage of this set of systems.Moreover,this set is able to capture the dynamic transformation failure process of the entire period of dynamic load effects,thus providing a more effective means of measurement for the construction of a dynamic constitutive model of soil and a dynamic failure process simulation.