An Extraction and Classification Algorithm for Concrete Cracks Based on Machine Vision

To solve the problem of large errors in extraction and the difficulty in classifying crack images in health monitoring of civil engineering structures, a new classification algorithm of concrete crack extraction based on machine vision is proposed in this paper. First, the gray difference between the image and the background is expanded by an adaptive nonlinear grayscale transformation. The improved OTSU threshold segmentation is used to extract the cracks, and the fracture points in the extracted results are connected by combining the extension direction of the fracture skeleton line and the gray feature of the crack edge to obtain the complete crack image. At the same time, the number of bifurcation points of the fracture skeleton line is calculated, a gray projection histogram of X axis and Y axis is obtained. Then, the classification characteristics of the cracks, such as the peak ratio of the gray histogram, the distribution ratio of the projection interval, and the mean square deviation ratio of the gray histogram are calculated. The obtained features are used as input to train a support vector machine classifier, which is then used to perform crack classification. The results of a simulation show that the proposed algorithm can extract crack images completely and precisely and can quickly and accurately classify the various types of cracks; thus, it has a good detection ability.