一种用于位移场分析的应变模式频域数字图像相关方法

Abstract

AbstractA spectral image correlation method considering the effect of displacement gradients upon displacement analysis in the Fourier frequency domain is proposed. A spectral image correlation criterion is studied and established to match the target subset image with the reference subset image in the frequency domain. An iterative algorithm for calculating displacement and strain is then derived, and the Hessian matrix remains unchanged within an iterative process. During the formation of the Hessian matrix, the greyscale gradients of the reference subset image can be calculated using the fast Fourier transform with high computational efficiency and accuracy. An accurate Fourier transform resampling technique, instead of the traditional interpolation method, is used to reconstruct or up-date the target image at a subpixel position in the deformed subset image. To verify the validity and accuracy of the proposed method, a series of images (2D-Challenge 1.0, sample 6) provided by the International DIC Challenge Committee (IDCC) was used for displacement analysis. The analyzed error results show that the proposed method is better than the published literature in accuracy. The second simulation experimental result with a strain of 50% is performed and the corresponding displacement accuracy is about 0.0032 pixel under the condition of 41 × 41 subset. Finally, a 14.85% uniaxial tensile test of a silicone rubber specimen is performed, and the analyzed displacement and strain distribution are given.