Tunnel Damage Detection And Research On Image Interpretation Method Based On SRC

Along with the development of national traffic and urban underground transportation infrastructure, lead to the size and the number of highway, railway and subway tunnel are on the rise. But the long-term operation of the tunnel will cause sorts of damages appear in the tunnel lining(metro tunnel grouting layer), such as cavity, faults, or cracks, etc. In order to ensure overall quality of the tunnel and eliminate operation safe hidden trouble, we need to implement damage detection about the tunnel. Ground penetrating radar(GPR) has the characteristics of fast, nondestructive and real-time imaging, has been widely applied in the nondestructive testing of transportation infrastructure. But GPR is electromagnetic detection equipment, can't directly reflects the characteristics of area target for imaging. Now expert experience is mainly used to interpret GPR images, but this method exist the shortcomings of inconsistencies explain results and long explain cycle. A kind of ground penetrating radar(GPR) data automatic interpretation method need to be invented urgently, to automatically locate and identify internal damages in the concealed work. The main contents include:We through the example of metro tunnel to introduce tunnel detection method and artificial interpretation of the radar image based on ground penetrating radar(GPR). grouting effect is evaluated by detecting grouting layer thickness and its internal distribution of damages. The dielectric constant of segment and the grouting body obtained by laboratory technology are assist to achieve the goal.The simulation data and the measured data of highway tunnel after clutter suppression are Implemented by sample classification. The time-frequency transform was carried out on the training sample. Then we change the time-frequency distribution matrices into a column vector, which can be used to build the time-frequency(redundant) dictionary. Then extracting the characteristics of training sample, such as large amplitude, energy, time and frequency entropy etc. The characteristics can be use to the classification of Support vector machine(SVM). At the same time these eigenvalues can constitute a feature dictionary for sparse decomposition classification.Sparse representation-based classifier is constructed to classify the damage of highway tunnel: Based on the time-frequency dictionaries, OMP method and its deformation SAMP are used in this article for sparse decomposition. And we compare the classification results. Support vector network can be obtained by input the extraction characteristic value to SVM, then result of SVM classification is compared with sparse classification results. To achieve sparse classification based on the characteristics dictionary, and compared its result with classification effect based on the time-frequency dictionaries. Finally, the depth of the damage areas position are located.