The Application Of Digital Image Processing In The Identification And Detection Of Cracks

The rapid-developmented economy promoted a large number of infrastructure’sconstruction, such as roads’ and bridges’, etc. As the time passing by, cracks will emerge onsurface of these constructions, which will either affect the aesthetics of this buildings, or evenworse, defect their safely use. In order to ensure these buildings against accidents from cracks,the regular detection and maintain are necessary. In this paper, digital image processingtechnology is applied on live identification and detection of the crack images.Crack identification is to discriminate whether an image contains cracks or not. Based onthe study of the pavement images, in the first part of this paper, a new algorithm of crackidentification is proposed, which is named as the crack identification algorithm based onequilibrium. This algorithm derives the cracks’ information by image-dividing, binarizing,and connected-component-picking, according to the characteristic that variation of crack localgray value is bigger while the background’s is smaller. Firstly, smaller images are obtained byimage-dividing, and then their variances are formulated. According to these variances, imageswith crack information are reserved. If there are cracks in original image, the smaller imagesbeen reserved display a continued line appearance, otherwise they will locate discretely.Secondly, simplified binarized image is obtained by reserved images, and furthermore, thelargest connected component is picked. Lastly, it is discriminated that the cracks do exist inimage or not and their trend by the length of largest connected component. By contrastexperiments, this algorithm is superior to the literature method in computing time andaccuracy.Crack detection is to extract unbroken crack information. In the second part of this paper,the crack-detection methods available are studied and improved, and a new algorithm isproposed on the basis of equilibrium. At first, threshold value method determined by thehistogram valley is stated to be the best one that is obtained by comparing the running timeand the binarization of several adaptive threshold segmentation algorithms. Then, imagedenoising is used in binarized crack image. For the rest of pseudo-cracks, spectral clusteringmethod is firstly used to obtain the unbroken crack. Finally, a new crack detection algorithm is proposed, which is named as the crack detection algorithm based on equilibrium. Thealgorithm can get rid of much noise, has fast running speed, and can be applied to the concretepavement as well as asphalt pavement crack detection.