Pavement crack detection based on two-scale clustering algorithm and 3D data

In order to identify the pavement cracks more accurately and efficiently, a kind of two-scale clustering pavement crack recognition method is proposed based on 3D data. First, the 3D data of pavement cracks are pretreated by means of the median filtering and the Otsu threshold segmentation algorithm, and the binary pavement crack images are thus obtained. Next, the irregular pavement crack area is characterized by a regular elliptical crack model as the basic unit, which represents a cluster basis. Then, the two-scale optimization criterion function is used to identify the basic units from the viewpoints of the distance and the angle deviation. Finally, the center of the final position of the cluster is determined based on the lever principle, and the area of the clustered cracks is characterized by using the minimum external ellipse of a complete crack. Moreover, the degree of circularity factors are adopted to judge the types of pavement cracks, and the parameters relevant to the damage extent of the road with reticular cracks are analyzed, which provides reference factors for the quantitative analysis of the road damage extent. The experimental results of the actual pavement cracks show that the proposed method is of a high degree of accuracy in detecting pavement cracks. ©, 2015, South China University of Technology. All right reserved.