A hybrid SURF-DIC algorithm to estimate local displacements in structures using low-cost conventional cameras. (October 2019)

• Record Type: Journal Article
• Authors: Cabo, C.; Ordóñez, C.; Muñiz-Calvente, M.; Lozano, M.; Ismael, G.
• Is Part Of: Engineering failure analysis. Volume 104(2019)
• Page Start: 807
• Page End: 815
• Publication Date: 2019-10
• Subjects: Image correlation -- Feature matching -- Displacements; System failures (Engineering) -- Periodicals; Fracture mechanics -- Periodicals; Reliability (Engineering) -- Periodicals; Pannes -- Périodiques; Rupture, Mécanique de la -- Périodiques; Fiabilité -- Périodiques; Fracture mechanics; Reliability (Engineering); System failures (Engineering); Periodicals; Electronic journals; 620.112
• DOI: 10.1016/j.engfailanal.2019.06.083
• Languages: English
• ISSNs: 1350-6307

Abstract

Abstract: A fast, efficient and low-cost method for the accurate determination of local displacements in structures based on computer vision techniques is presented in this study. First, displacements at pixel level are estimated using the SURF (Speeded-Up Robust Features) image matching algorithm, which detects corresponding points in the images. Then, the initial displacement estimations are refined at subpixel scale through image cross-correlation in the frequency domain. The image acquisition is simple and fast, using a conventional low-cost digital camera. The method was validated comparing the results obtained with those provided by a high performance system (ARAMIS 5 M) that integrates two cameras and image processing software. Results show that the proposed method is able to detect displacements with a precision of approximately 60% of the pixel size. The algorithm was implemented in a computer program in order to automatize the procedure. Highlights: Estimation of displacements in structures by using a standard digital camera and DIC. Development and use of algorithms of image matching and cross-correlation. The hybrid SURF-DIC algorithm developed allows obtaining a precision around 60% of the pixel size. A precision of 0.03 mm at a distance camera-object of 80 cm has been reached. Comparison of the results with that obtained by commercial packages (ARAMIS-GOM).