Concrete Crack Detection Algorithm Research Based On Digital Image

The increasing scale of social progress, economic development, infrastructureconstruction, concrete high-strength, high-tech building materials has been used in variety ofengineering. However, due to various reasons, the concrete structure will produce cracks andaffect its use, durability and aesthetics. A large number of experiments show that the crack inthe concrete structure is inevitable. When the crack width a certain value, it will affect thesecurity of the building. Therefore, we must attach great importance to the cracks of theconcrete, to control the extent permitted by. In this paper, the application of digital imageprocessing technology to detect cracks in the concrete surfaceCrack image binarization is to first give the cracks grayscale image to select anappropriate threshold, then this threshold to distinguish the pixels belonging to the targetcrack area and the background area pixels, the more obvious black and white images of thefracture characteristics, where the white portion represents the background region, the blackportion indicates the target crack region. Binary crack image, reducing the amount of imagedate, the crack image easier highlights the contours of the crack area, is conducive to furtherextract the edge of the crack. We propose a new the binarization algorithm-random movingalgorithm. The algorithm regarded an image as a concave-convex ground with "pits" and"ditches", where the "pits" represent the noise pixels and the "ditches" represent the crackpixels. First, this algorithm produces a certain number of seeds and the seeds’ position israndom. Then let every seed move in random and under a given rule until all of these seedsfall into the "pits" or "ditches". Finally, the locations of these seeds are marked, and then theoriginal image can be changed into a binary image which contains only "pits" and "ditches".The detection of cracks is extracted cracks in the image after binarization. This paperfirst introduces several classical edge detection algorithm, then test analysis for eachalgorithm. Second, we propose a new crack detection algorithm Random Moving Algorithm,this algorithm based on random moving binarization algorithm, the noise pixels in the binaryimages are removed by connected component analysis method and the region of the crack pixels can be located preciously. The experimental results show that: the proposed algorithmhas a strong stability to the nonuniform lighting conditions, and its performance is better thanthe classical Sobel and Canny edge detection method.