LAUSR

The crack is an important index to evaluate the strength of buildings. However, for the tiny cracks with low signal-to-noise ratio, traditional methods cannot obtain good detection results. This paper proposes a new algorithm for crack extraction based on improved tensor voting. On the crack images after preprocessing, firstly, a contour dilation and filtration is proposed for denoising. Then, the tensor voting algorithm is used to obtain the probability map of cracks. Finally, based on the probability maps, the real cracks are extracted successively through sampling, refining, center line tracking, and projected positioning. The experimental results show that the proposed method is robust to noise and has good results on crack extraction. It can effectively extract linear cracks with tiny size, low contrast and poor continuity.