Abstract:Considering the low contrast and severe noise features of wall crack images, this paper proposes a crack detection algorithm based on tensor voting and target tracking. On the basis of image binarization, the idea of clustering with tensor voting is used to highlight the linear target of the image. Then feature points are chosen on the highlighted linear target, and locations of cracks on the image are obtained through feature points tracking. Finally, the concept of re-cast is utilized to capture the original information of cracks. Test results show that this algorithm can detect cracks correctly on wall images and has high detection rate along with some degree of universality.