IMAGE ANALYSIS OF CRACKS IN CONCRETE: METHODOLOGY, OPPORTUNITIES AND PITFALLS
Price
Free (open access)
Volume
Volume 6 (2011), Issue 2
Pages
16
Page Range
145 - 161
Paper DOI
10.2495/DNE-V6-N2-145-161
Copyright
WIT Press
Author(s)
P. STROEVEN & H. HE
Abstract
Damage in concrete is mostly visualised in sections or at the surface of specimens subjected to internal and/or external loading. It continuously develops by growth and coalescence of tiny cracks into a spatial network structure. This structure can be seen as the finger-print of the material reflecting its history of loadings under given environmental conditions. The methodology of contrast improvement as an essential link in visualising damage is touched upon. However, a major focus of the paper is on describing damage by submitting images of crack patterns consisting of lineal features in the plane to a sweeping test line system and counting intersections. To obtain three-dimensional damage information in an economic way, the damage structure is assumed in the most general case revealing a partial orientation of mixed lineal and planar nature (the so-called ‘Stroeven-concept’). The practical cases are elaborated of prevailing compressive and tensile stresses. This reduces the number of unknown crack portions to two. As a consequence, quantitatively analyzing the image patterns can be restricted to vertical sections only. This involves a dramatic reduction of sawing efforts and simplifies the image analysis stage as well, of course. Only two orthogonal intersection counting operations are required for the assessment of specific crack surface area and of the degree and direction of preferred crack orientation. When observations would have been obtained in more directions, so-called roses of intersections (or intersection densities) can be determined. For very large images this would be circles. For random cracks in the image plane a circle around the origin, for oriented cracks, a circle through the origin is found. This concept, in addition to mathematical formulations is employed to demonstrate that automation of quantitative image analysis generally yields biased information, unless the analysis is executed under conditions discussed herein.
Keywords
Analog and digitised images, automation of image analysis, concrete, cracking, line scanning, rose of intersections