Partitioning Threshold Segmentation of Pavement Images Based on Combination of P-Tile and Histogram-Based Fuzzy C-Means Clustering

摘要

The technology of pavement cracks automatic detection has importance of the development of the China pavement traffic. The images segmentation of pavement is the key to the pavement images processing steps. Because the noise have a great influence on the pavement images, the traditional fuzzy C-means clustering (FCM) algorithm to the image segmentation have not satisfied results. In this paper, partitioning threshold is used. The global threshold is determined by P-tile algorithm and Histogram-based fuzzy C-means clustering algorithm; The local threshold is determined by the global threshold, P-tile algorithm and the feather of the images. For one thing the algorithm overcomes large amount of computing and the shortcomings of the slow speed of the traditional FCM; for another thing it reduces the scope of segmentation, and enhances the segmentation effectiveness; further more different subgraphs have different thresholds, so it avoids the defect of the global threshold, and make the image segmentation more accurate. The experimental results show that using algorithm in this paper it can better segment pavement cracks in the pavement images.