Research On Tunnel Crack Image Registration And Variation Trend Detection Algorithm

Crack is one of the important factors affecting the internal security of the tunnel.As time goes on,cracks may become wider than before,eventually seriously affecting the safety of subway operation.Therefore,specific dangerous cracks should be monitored closely.Because of outside environment and the capturing device,comparing images with same cracks taken at different times,the position and rotation of cracks will be different.This thesis proposed a tunnel crack image registration and variation trend detection algorithm based on image processing.This algorithm can match tunnel crack images,recognize images containing specific cracks through image databases and compare cracks' width variation.Mask Dodging is used to balance the illumination of images and enhance the contrast of cracks.OTSU method is used for threshold segmentation.Then a two-step filtering algorithm based on connected region is proposed to remove noises and achieve binary images only containing crack information.Thinning and debranching algorithm is used to extract the skeleton of the crack.In order to extract feature of crack image for the follow image registration,crack image feature extraction algorithm is proposed in this thesis.Improved Harris detector is used to extract feature point set of crack-skeleton image and then the set should be extended and normalized using proposed algorithm.Next,feature point sets of crack images should be matched.A registration algorithm based on Frechet-distance and Hausdorff-distance is proposed in this thesis to match two point sets.When detecting cracks' width variation,a width calculation method combines binary image and gray image is used.A principle called 1-? is used and the width of cracks is calculated in the pixel domain.Simulating cracks' width variation using MATLAB,the accuracy of the width calculation method is verified.At last,using proposed crack image registration algorithm,crack images collected on wall,road and tunnel are tested,showing that the registration algorithm works well in three scenes.In the pavement and subway tunnel crack image matching experiment,the false detection rates were 8%and 19% respectively.