Detection of Road Cracks with Multiple Images

Extracting the defects of the road pavement in images is difficult and, most of the time, one image is used alone. The difficulties of this task are: illumination changes, objects on the road, artefacts due to the dynamic acquisition. In this work, we try to solve some of these problems by using acquisitions from different points of view. In consequence, we present a new methodology based on these steps : the detection of defects in each image, the matching of the images and the merging of the different extractions. We show the increase in performances and more particularly how the false detections are reduced.