Three-dimensional Deformation Measurement Based On Digital Speckle Correction Algorithm

With the popularization of industrial intelligence,three-dimensional measurement system is widely used.Three dimensional sensing technology provides reliable technical support for industrial intelligence.As a non-contact three-dimensional sensing technology,digital speckle correlation measurement technology is more flexible than other measurement methods and can be used for fast full field measurement.However,the cost of commercial measurement system is high and the algorithm is not open source.In this paper,a low-cost digital speckle measurement system is designed.Considering the cost,algorithm complexity and convenience,a3 D speckle measurement scheme based on binocular vision is adopted.In this project,we use digital speckle correlation matching method to design the corresponding algorithm to measure the three-dimensional deformation of the object,and develop the corresponding test and analysis software,which can realize the independent and convenient measurement.The main research work of this paper is as follows:This paper summarizes and analyzes the advantages of digital speckle measurement algorithm compared with other optical measurement methods,compares the advantages and disadvantages of different digital speckle measurement systems,and determines the 3D speckle measurement scheme based on binocular vision from the aspects of algorithm complexity,experimental device cost and flexibility.By manually adding speckle texture to the surface of the specimen,the problem of difficult measurement of binocular vision weak feature object is solved.The mathematical model and error analysis of digital speckle are given.The digital speckle generation algorithm and corresponding quality evaluation algorithm are compiled.Several different types of speckle patterns are constructed to verify the feasibility of the algorithm.The optimal size of speckle is 2-4 pixels and the duty cycle of speckle is 0.8-0.9.In order to solve the problem of speckle image matching,four sub-pixel estimation algorithms and two integer pixel search algorithms are proposed for the existing matching algorithms,and their simulation experiments are analyzed.The results show that the sub-pixel estimation accuracy is affected by the size of the sub-pixel,the size of the speckle element and the form of the shape function.The average absolute error of the four Sub-pixel Search Algorithms is less than 0.005 pixel in the case of translation,and the higher-order shape function fitting accuracy is higher in the case of sinusoidal displacement,The inverse Newton optimization algorithm based on Gaussian window can reduce the error caused by improper shape function selection.Two integer pixel search algorithms are: Based on SIFT flow optical flow algorithm,used for small displacement image initial value calculation,can calculate the whole field displacement initial value,combined with IC-GN2 algorithm,the average absolute error is 0.006 pixels;Based on fast dense feature matching algorithm for binocular image matching.In order to solve the dense matching problem of binocular image,FDM-DSCM algorithm is designed,which combines IC-GN1 algorithm,fast dense feature matching algorithm and neighborhood seed point propagation algorithm,and can quickly calculate the pixel matching of the area to be tested.In order to calculate the3 D coordinates of the object before and after deformation,3D-DSCM algorithm is proposed,which is based on SIFT flow algorithm combined with four sub-pixel estimation algorithms to calculate the left and left image matching,and based on FDM algorithm and triangulation principle to calculate the 3D coordinates of the corresponding point before and after deformation.The 3D-DSCM algorithm is used to measure the rigid displacement and elastic deformation of the object.Finally,a convenient and practical software interface is designed,which integrates speckle image generation,image acquisition,two-dimensional image matching,three-dimensional deformation measurement and display into the software.In the experiment of complex topography measurement,the feasibility of the algorithm is verified;In the experiment of three-dimensional translation measurement,the average absolute error of forward displacement is 0.03 mm.