Research On Surface Fitting Of Correlation Coefficient Algorithm For Sub-Pixel Displacement Measurement In Digital Image Correlation Method

The digital image correlation method is a non-contact optical measurement method with simple equipment and high full-field measurement accuracy.It is widely used in many fields such as material detection,mechanical measurement and medical biology.With the continuous improvement of measurement performance requirements in various application fields,higher requirements have been placed on the measurement accuracy and real-time performance of digital image related methods.In the digital image correlation measurement algorithm,the iterative algorithm has a large and complicated operation amount,and the real-time performance is relatively poor in practical engineering applications.The surface fitting algorithm has the advantages of simple calculation,easy implementation,and good real-time performance,and has good prospects in practical engineering applications.In order to obtain the effect of different fitting functions on the accuracy and speed of the surface fitting algorithm and further improve the performance of the algorithm,based on the in-depth study of the surface fitting algorithm composed of different fitting functions,the BP neural The surface fitting algorithm of the network,and the genetic algorithm is used to optimize the algorithm.In order to systematically study the performance of the surface fitting algorithm composed of different fitting functions,this paper defines the algorithm as iterative surface fitting algorithm and non-linear algorithm according to whether the subpixel displacement solution method is different,that is,whether the analytical solution can be obtained directly.Iterative surface fitting algorithm.Aiming at the calculation accuracy,efficiency and adaptability to speckle noise of the algorithm,the numerical deformation experiments of real speckle images are used to compare the two types of surface fitting algorithms.The research results show that,among the non-iterative surface fitting algorithms,the binary Gaussian surface fitting algorithm has the best performance,and its mean error and standard deviation range are [-0.02,0.02] and [0,0.024],respectively.The time average is1.6s;Among the iterative surface fitting algorithms,the binary cubic polynomial surface fitting algorithm works best.The mean error and standard deviation range are [-0.02,0.015]and [0,0.003],respectively.The average calculation time is 3s.Based on the two types of algorithms,the calculation accuracy of the binary cubic polynomial surface fitting algorithm is higher,but its calculation efficiency is lower than that of the binary Gaussiansurface fitting algorithm.In order to further improve the calculation efficiency of the binary cubic polynomial surface fitting algorithm,the binary cubic polynomial surface fitting algorithm is combined with the neural network algorithm to establish a binary cubic polynomial surface fitting algorithm based on BP neural network.Through the non-mapping relationship between the input and output of the BP neural network,the correlation coefficient matrix in the surface fitting algorithm corresponds to the deformation displacement of the speckle pattern,thereby improving the calculation efficiency of the algorithm.The key parameter variables in the algorithm are analyzed and selected,and the algorithm model is constructed at the same time.The algorithm is compared with the original binary cubic polynomial surface fitting algorithm.The experimental results show that the algorithm is higher in calculation accuracy and efficiency than the original binary cubic polynomial surface fitting algorithm.In view of the shortcomings of the random initialization of the weight threshold of the BP neural network in this paper,using the global search feature of the genetic algorithm,the surface fitting algorithm of the BP neural network based on the optimization of the genetic algorithm is established,which effectively prevents the BP neural network from converging to the local optimal The phenomenon.The algorithm is compared with the unoptimized surface fitting algorithm based on BP neural network.Experimental results show that the optimized algorithm has higher calculation accuracy than the unoptimized algorithm,which verifies the effectiveness of the algorithm.In order to further verify the performance of the algorithm in this paper,a tensile test of the mechanical properties of materials was conducted.The experimental results show that the surface fitting algorithm based on genetically optimized BP neural network proposed in this paper is in good agreement with the measurement data of the extensometer and meets the measurement requirements.