Fast and Robust Digital Image Spearman's Rho Correlation for Displacement Measurement

Abstract

The robustness and computational efficiency of digital image correlation (DIC) are two key influencing factors for displacement field measurement applications. Especially when the speckle images are contaminated by salt-and-pepper noise, it is difficult to obtain reliable measurement results using traditional DIC methods. Digital image Spearman’s Rho Correlation (DISRC), as a new DIC technique, has certain robustness to salt-and-pepper noise, but incurs a high computational load when computing subset ranks. It is found that the DISRC can tolerate up to 15% noise level theoretically by analyzing the mean character of Spearman’s Rho. Meanwhile a fast scheme is proposed in which parallelization is adopted for precomputing subset rank and computing for displacement field to accelerate the DISRC. The simulation results indicate that the fast DISRC is about 60 times faster than the original one, and the displacement field results are almost the same between them. The DISRC not only gives as well results as zero-mean normalized cross-correlation (ZNCC) without any noise, but also can tolerate 20% noise level in simulations. A case study also verifies that the result by DISRC is better than ZNCC when contaminated by smaller amounts of noise. The conclusion is that the DISRC is a strong anti-interference DIC technique, which is very important in application under complex environment, and the fast scheme is an effective way to accelerate the DISRC.