The Study On Tunnel Crack Detection Method Using Image Processing Technology

Cracks as a common tunnel disease,seriously affecting the safety of the tunnel,so the regular detection of tunnel cracks is particularly important.The internal environment of the tunnel is complicated,and there are a lot of interference factors,which leads to problems such as uneven illumination,low target contrast,complex texture and noise.At present,the theoretical research on crack detection at home and abroad is mainly applied to industrial products,highway pavement,bridge engineering,and does not apply to the detection of tunnel cracks.In this paper,a tunnel crack detection method based on image processing technology is proposed,which can accurately identify and classify cracks,and calculate the geometric eigenvalues of cracks in the pixel domain.The main contents of the thesis are as follows:(1)A two-step preprocessing step is designed for the crack image.The gray scale of the image is adjusted by histogram equalization to adjust the gray scale of the image.Then the wavelet transform is used to enhance the crack image.Because of the advantages of directional selectivity and multiresolution analysis of wavelet transform,it is possible to establish a good correlation between the original image and the transform coefficient.Therefore,by separately processing the wavelet coefficients of different resolution levels,it is possible to achieve the target purpose.(2)A threshold segmentation algorithm based on maximum interclass variance method is designed.The gray value of the crack area is low,the gray value of the background area is high,and the image pixels are classified by the difference of the gray values of the two regions.The algorithm automatically determines the threshold by maximizing the variance between classes and corrects the threshold according to the segmentation effect.The experimental results show that the modified threshold can segment the crack from the background of the image and protect the crack information to the maximum extent,which makes up the shortcomings of the edge detection operator to noise sensitivity.(3)A fractal classification recognition algorithm based on BP neural network is designed.The four main features of the total number of cracks,the aspect ratio of the minimum circumscribed rectangle,and the projection in the horizontal and vertical directions are taken as input vectors of the BP neural network.The output results include transverse cracks,longitudinal cracks,diagonal cracks and reticular cracks Four types of cracks.A total of 83 images with various types of cracks were selected as training samples.The BP neural network was trained and the crack images were identified by BP neural network.The statistical results showed that the recognition rate of the four cracks was More than 90 percent.(4)Calculate the crack image obtained by different algorithms based on the calculation method of the geometric eigenvalue of the cracks,calculate the area of the irregular cracks and the length and width of the regular cracks.The results verify the feasibility and accuracy of the crack treatment method Which provides a reliable basis for the evaluation of the level of crack disease.