Accelerating parallel digital image correlation computation with feature mesh interpolation

Abstract

Parallel digital image correlation (DIC) computation has been increasingly used in DIC community due to the growing requirement on real-time deformation monitoring. In existing parallel DIC speeded up with feature-based initialization method, high parallelism requires more valid initialized points, which means more feature points are needed. However, more feature points significantly increase the time of feature extracting and matching, thus reducing the whole parallelism. To solve this problem, we propose to construct a mesh of matched feature points and assign initial values by mesh interpolation. The proposed method depends less on the feature distribution and provides more and better valid initialized points with much less feature points. Along with the well-designed parallel process, the proposed method is parallel-friendly, scalable and can realize high parallelism. The effectiveness, parallel performance, and the ability of real-time measurement of the proposed method are demonstrated by processing real experimental images.