Digital Image Correlation and Edge Detection: Applications in Materials Testing

Digital image processing (DIP) methods have been applied in civil engineering for many years. In this dissertation, DIP methods are applied to (1) quantify the mechanical properties of metallic and cementitious composites loaded in uni-axial tension, and (2) evaluate the effectiveness of removing paint from concrete surfaces. DIP is an optical method to characterize the behavior of the visible domain on the surface of the specimen. Strain, an important measurement in mechanics of materials, can be measured by a digital image processing method known as Digital Image Correlation (DIC). Traditional DIC tracks the change in position of speckles on the surface of a specimen to determine the surface strain. In this dissertation, a novel pattern, a painted rectangle on the specimen surface, is developed to measure strain. This approach uses an edge detection method to track changes in the length of the rectangle. It is a target-based deformation method, which explicitly considers the deformation of the target (rectangle) to calculate strain. By computing the change in length of the rectangle in successive images, the strains are determined. Both pixel-level and subpixel-level accuracy levels for different accuracy requirements are studied. The pixel-level is implemented using ImageJ software, while the subpixel-level is accomplished using two algorithms: Gaussian fitting and spline interpolation, both implemented using Matlab. The two digital image methods (DIC speckles and rectangle edge detection) are compared with laser extensometry and strain gage measurements for uni-axial tension tests of A36 steel specimens. The strain measurement results are comparable to the laser extensomety and strain gages if high quality images are captured. DIP methods only characterize the behavior of the visible domain on the surface of the specimen. A combination of technologies (both traditional and emerging) may be necessary to fully characterize the mechanical response of a material or structure subjected to loading. Two digital image processing methods are established to evaluate the effectiveness of cleaning paint from concrete surfaces. The principle of this evaluation is to select a suitable gray intensity threshold value for each paint color to distinguish it from the concrete color. One method is histogram-based and selects the threshold value based on the histograms of the pixel values of the paint and concrete. The other method is an edge-based method, which selects the gray intensity of the edge as the threshold value. Based on the results from the binarized images produced by these two methods, the histogram-based method is more suitable for the darker paints and the edge-based method is more suitable for the white paint. These methods could also be applied to evaluate the removal of corrosion by-products from steel or even stains from teeth. Five strain measurement methods are applied to measure the deformation of fabric reinforced cementitious matrix (FRCM) composites tested in uniaxial tension. The five methods include: laser extensometry, clip-on extensometry, strain gages, and two digital image methods (DIC speckle and rectangle edge detection). The strain gages failed prematurely due to cracking of the FRCM. The two digital image methods can only analyze in-plane displacement. For out-of-plane displacement, a 3-D DIC method must be used