Automatic Road Cracks Detection and Characterization Based on Mean Shift

In view of practical problems such as road cracks damage detection with low efficiency,less security,and poor data standardization,an unsupervised method for the automatic recognition and characteristic measurement of road cracks is proposed.Firstly,apply the image to be segmented based on the method called split-and-merge.To be more exact;split the image into smaller blocks,after iterative smoothing and segmentation with the MS method,merge these blocks together.Secondly,two different strategies,both based on the MS method,are used to directionally trace the pixels to extract the skeleton of cracks.And the complete cracks are obtained from the interpolation on the skeleton.Meanwhile,calculate the parameters,the morphological characteristics of cracks are measured.The experimental results show that the proposed algorithm can effectively identify and accurately measure the cracks.