Extending Digital Image Correlation to Reconstruct Displacement and Strain Fields around Discontinuities in Geomechanical Structures under Deformation Ghulam Mubashar Hassan, Cara MacNish School of Computer Science & Software Engineering, University of Western Australia ghulam.hassan@research.uwa.edu.au, cara.macnish@uwa.edu.au Arcady Dyskin School of Civil & Resource Engineering, University of Western Australia arcady.dyskin@uwa.edu.au Abstract Reconstruction of displacement and strain fields in ge- omechanical structures from surface images is a challeng- ing task. Digital Image Correlation (DIC) is a well known technique to achieve these tasks if deformation is contin- uous but it fails in the presence of discontinuities. This paper investigates the application of the DIC technique to displacement and strain field reconstruction in the presence of discontinuities, and presents a post-processing algorithm that leverages the convergence results in DIC to reconstruct displacement and strain fields around discontinuities with high accuracy. The proposed algorithm uses the results ob- tained from DIC and concentrates on the area where DIC fails. Pattern matching is conducted on the area around the discontinuities and associated displacement is found for each pixel. The proposed algorithm is tested using two dif- ferent discontinuity scenarios: dislocation and fracture in structures. The results show that the proposed algorithm successfully reconstructs the displacement and strain fields to subpixel accuracy of 1/10th of a pixel. 1. Introduction Monitoring deformation is an important task in the field of engineering and sciences, for example in civil and ma- terials engineering to reconstruct strain and displacement fields in structures, in electronic and computer engineering to measure thermal expansion of circuit components and boards, in biomechanics to measure deformation and strain in different parts of the body, in avionics to measure strain in load testing of aircraft and in seismic geology to measure vibrations or movement of tectonic plates [1]. For moni- toring the stability of structures, reconstruction of displace- ment and strain fields is necessary. This helps to predict approaching failure or instability in the structures and take appropriate actions to avoid disaster. The reconstruction of the strain field requires measur- ing strain at many points in the structure. Strain gauges are commonly used to do this task. A strain gauge is an electro- mechanical component which measures change in electrical resistance due to load and calculates the strain in a single di- mension at the point where it is affixed. The displacement measured at different points help to reconstruct the displace- ment and strain fields. Strain gauges provide high accuracy but have some limitations. Strain gauges measure the strain in a single direction only. To reconstruct the strain field, multiple strain gauges are required to measure the strain at different points in dif- ferent directions which is an expensive and labor intensive process. Furthermore, strain gauge installation may ob- struct the normal use of the structure, such as in excavation processes and geomechanical structures. Under these circumstances, photogrammetry with Digi- tal Image Correlation (DIC) may be used as a contactless technique to measure displacements and strains. DIC is an image processing technique which was introduced for mea- suring displacement in deformable solids [2]. Because de- formation in solids may be very small, it is required to pro- vide a high subpixel accuracy of 1/10th or 1/100th of a pixel in measuring displacements. DIC compares two images, reference and deformed, which are taken before and after deformation. The algo- rithm selects a pixel and its neighboring area, or subset, in the reference image. The subset is then correlated with sub- sets in the deformed image and the new position of the pixel in the deformed image is approximated. This process is re- peated for all pixels in the reference image and the displace- ment and strain fields are reconstructed. DIC is commonly used due to its relatively simple exper- imental setup and implementation. Many different variants 2015 IEEE Winter Conference on Applications of Computer Vision 978-1-4799-6683-7/15 $31.00 © 2015 IEEE DOI 10.1109/WACV.2015.100 710