Pavement crack detection based on improved tensor voting

Abstract

Conventional pavement crack detection algorithms can hardly detect pavement cracks accurately due to the intensity inhomogeneous and complicated noises over the pavement surface. In this paper, a novel pavement crack detection method based on tensor voting is proposed. Firstly, the improved Retinex algorithm is adopted to eliminate the effect of uneven lighting. Then, a crack enhancement algorithm based on saliency is presented. This is followed by Otsu thresholding to acquire the crack seeds. Motivated by the framework of tensor voting, we remove noises and connect the crack seeds to generate integrated cracks. Finally, real cracks are extracted through non-maxim suppression algorithm. The proposed method has been tested on a real pavement crack database collected through a Chinese highway survey. The experimental results demonstrated that this method is more accurate and robust than traditional algorithms.