EVALUATION OF CRACK WIDTH DISTRIBUTION CHARACTERISTICS OF RC WALL USING IMAGE PROCESSING

Abstract

 In evaluating the reparability of reinforced concrete building structures, it is very important to properly evaluate the crack amount (crack width, crack length). However, in actual seismic design, it is general to evaluate that the proof stress of the building structure exceeds the specified external force. Even if the seismic response is evaluated during the seismic design, seismic crack width control is not in an explicit form. In recent years, research has been continued toward acquisition of a database of damage amount of RC structural members that contributes to reparability evaluation. However, the measurement crack width has been classified according to the traditional damage level classification criteria because of the labor required to measure the total amount of cracks in experimental filed and the enormous amount of data to be collected. Therefore, in this study, we constructed a high-precision damage amount measurement system using an image processing and morphological operations, and we verified whether or not the number of cracks could be measured in detail for RC wall.

 Firstly, image process was performed to extract crack area from RC wall damage image, then it was utilized that the morphological operations proposed in this paper to calculate crack width distribution characteristics which was defined as the ratio of crack length with each crack width class to the total crack length. The morphological operations for damage measurement was divided into two parts; normal measurement and special measurement. Normal measurement was designed for calculating crack width which was at smooth crack contour as edge pixels. And the special measurement was designed for calculating crack width which was at corner of crack contour. Based on the crack width obtained from normal measurement and special measurement, linear interpolation to acquire crack width distribution was performed. Through the above-mentioned process, crack width distribution characteristics were available.

 Secondly, the crack width distribution characteristics was compared between RC wall and RC beam. The variance of the lognormal distribution simulating the crack width distribution characteristics of RC walls was smaller than that of RC beam members. It was implied that it was limited to one or two cracks that widen in RC beams, whereas more cracks were similarly widen in RC wall.

 Finally, verifying the image processing measurement method, the reason why the maximum crack width by image processing measurement method did not match to the visual measurement by human power was discussed. One reason was the difficulty of binarization threshold setting due to ambiguity of crack area boundary. Second reason was the counting the corner of crack. If the corner was accurately considered, the crack width becomes locally thick, but the locally thick portion is not recorded as the crack width by visual measurement. However, if the precise crack distribution characteristics are acquired, it is only necessary to delete the local crack width that is practically unnecessary. It is considered that the validity can be secured much more than fabrication of unmeasured data.