Evaluation of digital image correlation techniques using realistic ground truth speckle images

Abstract

Digital image correlation (DIC) has been acknowledged and widely used in recent years in the field of experimental mechanics as a contactless method for determining full field displacements and strains. Even though several sub-pixel motion estimation algorithms have been proposed in the literature, little is known about their accuracy and limitations in reproducing complex underlying motion fields occurring in real mechanical tests. This paper presents a new method for evaluating sub-pixel motion estimation algorithms using ground truth speckle images that are realistically warped using artificial motion fields that were obtained following two distinct approaches: in the first, the horizontal and vertical displacement fields are created according to theoretical formulas for the given type of experiment while the second approach constructs the displacements through radial basis function interpolation starting from real DIC results. The method is applied in the evaluation of five DIC algorithms with results indicating that the gradient-based DIC methods generally have a quality advantage when using small sized blocks and are a better choice for calculating very small displacements and strains. The Newton-Raphson is the overall best performing method with a notable quality advantage when large block sizes are employed and in experiments where large strain fields are of interest.

A three-frame digital image correlation (DIC) method for the measurement of small displacements and strains

“…All deformed images in the numerical experiments are obtained by deforming a common 512 × 512 pixel reference image using the RBF interpolation framework introduced in [33]. The reference image is part of a 2208 × 3000 pixel 8-bit greyscale frame captured with a Pixelink PL-A782 camera that shows an aluminium specimen.…”

Improved Newton–Raphson digital image correlation method for full-field displacement and strain calculation

“…Each scenario contains an image pair composed of a "reference" and a "deformed" image. The first two scenarios correspond to a "plate with hole under biaxial stress" model [20] with their deformed images numerically obtained by warping a common reference speckle image using known displacements and radial basis function image interpolation [9]. In third and fourth experimental scenarios which represent real experiments, a plastic film specimen 25.3 mm in width, 250 mm in length and 0.1 mm in thickness with a lateral slit into its right edge and an aluminum specimen 25.25 mm in width, 280 mm in length and 2.95 mm in thickness with two lateral slits on both edges undergo uniaxial load in a vertical upward direction.…”

“…Since their introduction [1][2][3][4] various classes of algorithms were developed, the most prominent of these involving the Newton-Raphson method of partial differential correction [5][6][7]. When compared to the other methods [8,9] it shows higher sub-pixel accuracy and allows the reliable use of the more complex linear and quadratic local motion models at the cost of increased computational complexity and sensitivity to the interpolation method used in the minimization process. The adoption of the method has grown due to its quality advantages despite the higher computational requirements.…”

Photoconductors for Measuring Speckle Motion to Monitor Ultrasonic Vibrations

“…These methods are generally used in low speed applications --CCD arrays with low read-out rates and long integration times (on the order of milliseconds). A recent report on different algorithms (Cofaru, 2010) finds that resolution is on the order of 10 -3 pixel widths, generally speaking around δ min = 10 nm. Although this is on the same order as the detection limit of the SOI sensor, the SOI photoconductive mesa sensor described here does so at 1 MHz bandwidth.…”