Improved digital image correlation for in-plane displacement measurement Asloob Ahmad Mudassar1,* and Saira Butt2 1
Pakistan Institute of Engineering and Applied Sciences (PIEAS), Nilore, Islamabad, Pakistan
2
Pakistan Institute of Nuclear Science and Technology (PINSTECH), Nilore, Islamabad, Pakistan *Corresponding author: [email protected] Received 1 August 2013; revised 25 October 2013; accepted 10 November 2013; posted 3 January 2014 (Doc. ID 195054); published 7 February 2014
Electronic speckle photography (ESP) for in-plane displacement (IPD) and deformation measurements is well known with its more modern form, digital image correlation (DIC). Two speckle images of an optically rough surface before and after deformation, called reference and test images, are recorded and processed for IPD or deformation measurement of the test image with respect to the reference image. The reliability of ESP in measurements depends strongly on the postprocessing of the two images by DIC, which we have referred to as conventional DIC. In this paper, we are proposing a small but useful modification in the existing DIC methods by introducing some additional steps, which drastically improves the results obtained with the existing techniques. The modification to the conventional DIC method has been referred to as modified DIC. Computer-simulated and experimental results have been presented to validate the superiority of modified DIC over conventional DIC methods. © 2014 Optical Society of America OCIS codes: (120.0120) Instrumentation, measurement, and metrology; (120.3940) Metrology; (120.6150) Speckle imaging; (100.2000) Digital image processing. http://dx.doi.org/10.1364/AO.53.000960
1. Introduction
Digital image correlation (DIC) has been an active research area since the past two decades and is well known in the measurement of in-plane displacement (IPD). Different parameters of the technique have been addressed in the literature for improvement of accuracies in the measurement results and for removal of different defects encountered during experimentation [1–53]. In DIC, two images called reference (before application of load) and test (after the application of load) of an optically rough surface under illumination by a light source (e.g., a laser beam) are recorded. These two speckle images are cross-correlated for measurements of in-plane deformations (displacements, rotations, strains) that the test image has undergone with respect to the 1559-128X/14/050960-11$15.00/0 © 2014 Optical Society of America 960
APPLIED OPTICS / Vol. 53, No. 5 / 10 February 2014
reference image. We are presenting a modified version of the conventional DIC method, which we shall be referring to as modified DIC. Accuracy in DIC is based on accurate recording of images. Before discussing the modified DIC technique, we want to highlight the issues that may affect the accuracy of displacement measurements. For the two images in DIC, a single illumination source and a single charged-coupled device (CCD) are required [1,2]. Straining an illuminated object produces deformation. Excessive straining (like excessive stretching of a spring about its elastic limit) of the object may result in speckle patterns de-correlated completely or partially with the speckle pattern in the un-deformed state of the object surface and the DIC would not work properly [3]. High-resolution microscopes can be used to record images and deformation of the order of micrometers to nanometers can be measured [4–11]. The recording process requires the sensor array of the CCD to
be held nearly parallel to the object’s planar surface [12,13]. System magnification is kept constant by either using a telecentric system or positioning the CCD at a relatively large distance from the specimen [12–14]. Speckles in microscopy-based images should be related linearly with object IPDs, otherwise geometric distortions must be taken into account for accurate measurements [4,7,15–18]. In case of homogeneous IPD, reference image in full dimension is used in the cross-correlation process to measure the bulk IPD; otherwise small square patches (subsets) from the reference image are used. Under large deformations, square patches are distorted and first-order and second-order shape functions may then be fitted to guess the shape of the deformed reference subset [20]. The sum of squared difference approach [19] is linearly related to DIC [21] and both require constant illumination and zero light offset. Use of plane waves or restriction to smaller IPD with diverging beams ensures constant illumination. Zero illumination offset can be assured if the light source intensity does not change during measurements [19–21]. Finite pixel size and pixel pitch affect DIC accuracy [22]. Subpixeling using a bi-cubic spline or a bi-quintic spline for accuracy improvement can be done either at the images before DIC or at the image after DIC, with respective improvement by a typical factor of 4 or 10 [23]. The size of the reference subset may vary from five to seven times the average speckle size and is critical when the deformed image changes its shape. Details of choosing a proper size subset are given in [24]. DIC is basically a spatial domain technique, but its Fourier domain version works equally well [25–27], but in the presence of negligibly small rotation (
Pakistan Institute of Engineering and Applied Sciences (PIEAS), Nilore, Islamabad, Pakistan
2
Pakistan Institute of Nuclear Science and Technology (PINSTECH), Nilore, Islamabad, Pakistan *Corresponding author: [email protected] Received 1 August 2013; revised 25 October 2013; accepted 10 November 2013; posted 3 January 2014 (Doc. ID 195054); published 7 February 2014
Electronic speckle photography (ESP) for in-plane displacement (IPD) and deformation measurements is well known with its more modern form, digital image correlation (DIC). Two speckle images of an optically rough surface before and after deformation, called reference and test images, are recorded and processed for IPD or deformation measurement of the test image with respect to the reference image. The reliability of ESP in measurements depends strongly on the postprocessing of the two images by DIC, which we have referred to as conventional DIC. In this paper, we are proposing a small but useful modification in the existing DIC methods by introducing some additional steps, which drastically improves the results obtained with the existing techniques. The modification to the conventional DIC method has been referred to as modified DIC. Computer-simulated and experimental results have been presented to validate the superiority of modified DIC over conventional DIC methods. © 2014 Optical Society of America OCIS codes: (120.0120) Instrumentation, measurement, and metrology; (120.3940) Metrology; (120.6150) Speckle imaging; (100.2000) Digital image processing. http://dx.doi.org/10.1364/AO.53.000960
1. Introduction
Digital image correlation (DIC) has been an active research area since the past two decades and is well known in the measurement of in-plane displacement (IPD). Different parameters of the technique have been addressed in the literature for improvement of accuracies in the measurement results and for removal of different defects encountered during experimentation [1–53]. In DIC, two images called reference (before application of load) and test (after the application of load) of an optically rough surface under illumination by a light source (e.g., a laser beam) are recorded. These two speckle images are cross-correlated for measurements of in-plane deformations (displacements, rotations, strains) that the test image has undergone with respect to the 1559-128X/14/050960-11$15.00/0 © 2014 Optical Society of America 960
APPLIED OPTICS / Vol. 53, No. 5 / 10 February 2014
reference image. We are presenting a modified version of the conventional DIC method, which we shall be referring to as modified DIC. Accuracy in DIC is based on accurate recording of images. Before discussing the modified DIC technique, we want to highlight the issues that may affect the accuracy of displacement measurements. For the two images in DIC, a single illumination source and a single charged-coupled device (CCD) are required [1,2]. Straining an illuminated object produces deformation. Excessive straining (like excessive stretching of a spring about its elastic limit) of the object may result in speckle patterns de-correlated completely or partially with the speckle pattern in the un-deformed state of the object surface and the DIC would not work properly [3]. High-resolution microscopes can be used to record images and deformation of the order of micrometers to nanometers can be measured [4–11]. The recording process requires the sensor array of the CCD to
be held nearly parallel to the object’s planar surface [12,13]. System magnification is kept constant by either using a telecentric system or positioning the CCD at a relatively large distance from the specimen [12–14]. Speckles in microscopy-based images should be related linearly with object IPDs, otherwise geometric distortions must be taken into account for accurate measurements [4,7,15–18]. In case of homogeneous IPD, reference image in full dimension is used in the cross-correlation process to measure the bulk IPD; otherwise small square patches (subsets) from the reference image are used. Under large deformations, square patches are distorted and first-order and second-order shape functions may then be fitted to guess the shape of the deformed reference subset [20]. The sum of squared difference approach [19] is linearly related to DIC [21] and both require constant illumination and zero light offset. Use of plane waves or restriction to smaller IPD with diverging beams ensures constant illumination. Zero illumination offset can be assured if the light source intensity does not change during measurements [19–21]. Finite pixel size and pixel pitch affect DIC accuracy [22]. Subpixeling using a bi-cubic spline or a bi-quintic spline for accuracy improvement can be done either at the images before DIC or at the image after DIC, with respective improvement by a typical factor of 4 or 10 [23]. The size of the reference subset may vary from five to seven times the average speckle size and is critical when the deformed image changes its shape. Details of choosing a proper size subset are given in [24]. DIC is basically a spatial domain technique, but its Fourier domain version works equally well [25–27], but in the presence of negligibly small rotation (
