Measurement of crack width from the two-dimensional sonar images of the moss-covered surface of underwater concrete components based on K-means clustering

Abstract

Imaging sonar can precisely measure the sizes of some large defects on the surface of underwater concrete components (UCCs). However, it is challenging for image sonar to measure the widths of small cracks on UCCs. Another difficulty is the coverage of aquatic organisms, such as moss. The moss coverage divided into four stages changes the hardness and roughness of UCCs to cause wrong signal intensity of echo, crack image generation, and crack width measurement. This article first illustrates the imaging principles of surficial cracks in UCCs using a two-dimensional imaging sonar and derives a theoretical equation of crack width measurement. Then, the appropriate settings of two main factors in actual inspection, scanning distance l and pitch angle α, are determined by experimentally obtaining the measurement error distribution of concrete cracks with various stages of moss coverage. It also summarizes the correction equation of crack width measurement at different stages of moss coverage. For the case of unknown stages of moss coverage, the peak gray P and the mean value M of fitting the Gaussian curve of the gray-level histogram of the sonar images are experimentally proved to be effective in distinguishing the different stages of moss coverage by a K-means clustering process. Experimental results indicate that 0.75 m ≤ l ≤ 2.0 m and 15° ≤ α ≤ 30° is the appropriate ranges of l and α in the practical measurement. The boundary line equation to separate Stage I and Stages II & III of moss coverage is 60 ≤ P + M ≤ 80. The boundary line equation to separate Stages II & III and Stage IV is 75 ≤ P + M ≤ 110.