Target-free Stereovision Approach Based On Structural Surface Features For 3D Deformation Measurement

For state evaluation and deformation measurement of common civil infrastructure,vision measurement,as one of the methods for nondestructive testing,can have many edges over contact techniques.However,traditional vision measurement methods require the application of artificial targets on the surface of the structure,it still has a lot of inconveniences and limitations.Extracting structural surface information and relying on these for vision measurement is a more practical idea.According to the defect of the artificial target-based method of determining in projects,this paper investigates vision measurement approaches based on features of the structural surface.This paper studies the algorithms of features extracting,image matching and 3D reconstruction.The main research contents include the following three parts:Firstly,a stereovision measurement approach based on feature points of the structural surface is presented,which applies the scale-invariant feature transform(SIFT)algorithm to stereovision measurement.By preprocessing the image sequence,extracting feature points of strctural surface,removing interference points,establishing reference feature point group and matching feature point set.The 3D displacement field of the structural surface can be obtained by matching feature points.The centroids of the point sets are matched to obtain the 3D point displacement of the structural surface.With reference to the finite element technique,the extracted keypoints are triangulated,and the 2D deformation of the structural surface can be obtained according to the geometric relationship.Then a 2D displacement field measurement approach and a 2D strain field measurement approach are proposed.A simulation of rigid-body translation measurement test,a material tensile deformation measurement test,and a 3D translational parallel mechanism measurement test were completed,and the test results indicate that the proposed approaches are stable.Secondly,the stereovision measurement approaches based on structural surface grayscale feature are introduced,including the methods based on structural surface high and low contrast contrast.When there are few obvious feature points on the structural surface,the deformation information can be analyzed by the contrast information.High contrast means that it can gather many orientation information,the orientation code matching method is applied to the vision measurement,by establishing and matching the orientation code histogram and the orientation code,a high-contrast measurement approach is proposed.In the case of low contrast requires contrast enhancement of the originally images,establishing and matching the representative value histogram and the binary image.On basis of the idea mentioned above,a low-contrast measurement approach is proposed.A simulation of a rigid body translation measurement test and a 3D translational parallel mechanism measurement test were completed,and the results reveal that the technique can achieve sub-pixel accuracy.Finally,combined with the the proposed approaches,softwares of structural 3D deformation measurement is developed.To verify that the proposed approaches are practical in the fields of structural engineering,a three-story frame model shaking table test was designed and implemented,and the model measured by the approaches based on surface feature points and contrast.It is found that the proposed method has outstanding performance in both the time domain and the frequency domain,which is in good agreement with the traditional method.A T-shaped wall structure shaking table test was completed,and the structure is measured by the approaches based on surface contrast.Compared with the conventional techniques,the displacement deformation of multiple measuring points can be given,and the structural dynamic property can be accessed.