Automatic Identification of Pavement Cracks Using Mathematic Morphology

According to the characteristics of pavement images, an innovatory pavement cracks identification algorithm based on mathematic morphological is proposed. The identification can be divided into three phases: By using Multiscale Alternating Sequential Filters operation, impulse type noise as well as Gaussian noise in pavement image is eliminated. Then the resulting "ideal" crack edges can be extracted by using multi-scale morphologic edge detection method. At last, the watershed transform via markers algorithm is used to segment cracks of pavement image. Compared with conventional methods, the results obtained from test images show that the modified morphological cracks identification methods are more effective.