Pothole detection on asphalt pavements from 2D-colour pothole images using fuzzy c-means clustering and morphological reconstruction

摘要: Abstract In general, potholes on asphalt pavements can be detected and represented in 2D 3D. However, pothole detections through 3D imaging image reconstructions have proven to expensive terms of acquisition equipment the computational processing requirements time. For at incipient formations, their detection, representation quantification surface-area are important for timely maintenance repairs. By casting pavement segmentation detection as a problem clustering multivariate features within mixed pixels (mixels), this study presents low-cost vision image-based approach road urban areas. The is based priori integration multiscale texture-based filtering textons using wavelet transform, into superpixel defects non-defects fuzzy c-means (FCM) algorithm. extraction extrema (minima maxima) hybrid wavelet-FCM results, fine morphological reconstruction adopted further smoothen recognize contour potholes. methodology implemented MATLAB prototype, tested validated 75 experimental datasets. With mean CPU run-time 95 seconds, average accuracies by comparing results manually segmented ground-truth data were determined Dice coefficient similarity, Jaccard Index sensitivity metric 87.5%, 77.7% 97.6% respectively. magnitudes standard deviation percentage errors size extractions 8.5% 4.9% show that with well-planned condition surveys, proposed algorithm suitable from images acquired consumer-grade sensors.