Multiple Disease Detection And Intelligent Identification Of Tunnel Lining Structure

In recent years,tunnel engineering construction in China has flourished,and a large number of tunnels have been put into operation.Due to long-term operation and complex environmental factors,tunnel lining structures are prone to cracks,cavities,and other diseases,posing certain safety hazards.Therefore,the detection and identification of tunnel lining structural diseases are indispensable for ensuring the safe operation of the tunnel.At present,traditional manual inspection methods and image processing detection methods cannot meet the requirements of tunnel disease detection and recognition due to low efficiency and strong subjectivity.In response to the above issues,this thesis takes tunnels during operation as the research object,and studies the detection and intelligent identification of surface cracks and internal diseases in tunnel lining structures.The main work of this thesis is as follows:(1)For the problem of detecting apparent cracks in tunnels,preprocessing and segmentation algorithms based on image processing were studied.The advantages and disadvantages of threshold segmentation and edge detection operators were compared and analyzed through experiments.A method combining iterative threshold segmentation and Canny operator was proposed to obtain crack targets.(2)For the classification and recognition of tunnel surface cracks,based on feature extraction combined with machine learning classifier methods,the extraction of HOG and LBP features was first studied.It was found that their feature vector dimensions were too large,and GLCM and SIFT features were used to compensate.Then,the extracted features were fed into a support vector machine for classification.Finally,the accuracy of different feature extraction methods was compared based on the classification effect.(3)For the classification and recognition of tunnel apparent cracks,based on the depth learning method,a tunnel apparent crack recognition algorithm based on transfer learning and integrated learning is proposed.Based on transfer learning training,three classification model algorithms,Goog Le Net,Res Net and Efficient Net,are compared.In order to improve the robustness and accuracy,the idea of integration is also cited.The classification results of the three networks are fused by voting.The results show that the accuracy of the algorithm in this thesis is better than that of the single convolutional neural network algorithm.(4)For the detection of internal diseases in tunnels,based on ground penetrating radar signal processing technology,three radar data denoising,clutter suppression,and imaging algorithms were studied,and the algorithm with the best detection effect was selected through simulation data.