|
Abstract Crack is one of the most common and serious defects in tunnel lining. In light of the existing problems of conventional image recognition methods，an automatic crack recognition method in tunnel lining based on characteristics of local grids in images is presented. A lining image is firstly divided into local grids of 8 Pixel×8 Pixel. Cross-shaped templates are designed based on the characteristics of luminance difference and contrast difference between different directions in local grids. The pixel with minimum gray value in each grid can be recognized as one potential crack seed by template calculation. Discrete crack seeds are finally linked together to form an intact and continuous crack cluster using seed linking algorithm. During the linking process，the direction，length and width of cracks are measured automatically. The optimal parameters and threshold of the proposed algorithm are estimated using receiver operating characteristics(ROC) curves. The reliability and accuracy are validated by means of qualitative and quantitative analyses. Application cases show that the proposed method can achieve good effects of crack recognition，especially for the lining images containing minor cracks and leakage；and the reliability and recognition rate are higher than those of other conventional image recognition methods.
|
|
Received: 25 November 2011
|
|
|
|
2012, Vol. 31 
