Automatic Road Crack Detection by Selection of Minimal Paths

Abstract

Automatic detection of road cracks from pavement images has become an important challenge in many countries. Among the different methods proposed in the literature, this paper proposes to combine shortest-paths previously estimated in the image. The proposed method takes account of both photometric and geometric characteristics of cracks simultaneously and requires a few informations a priori. It has been tested on image data sets collected by a dynamic imaging system.