The Research On Visual Mining Method Of Time-Varying Structural Modal Of GPR Data

Ground Penetrating Radar(GPR)is an electromagnetic exploration technology used to detect the distribution of underground structures.The data collected by GPR has the characteristics of large scale,complex spatiotemporal characteristics,and sensitivity to the survey environment.Among them,the time-varying structural mode is the most valuable information in GPR data and plays a very important role in understanding the distribution of underground structures.However,due to a large amount of noise and clutter in GPR data,there are some problems such as high time complexity and poor robustness in traditional time-varying structural mode mining methods.Therefore,improving the robustness of the time-varying structural modal mining process and reducing the time complexity has become one of the key issues to be solved in the GPR application field.Visual data mining technology can integrate theories,methods and tools in different fields,provide advanced analysis methods,interactive technologies and visual expressions,which can help people communicate quickly with massive and complex information.Therefore,based on the advantages of visual data mining techniques,we solve the problems of high time complexity and poor robustness in the process of time-varying structural modal mining and carry out the research of GPR data time-varying structure modal visual mining method.The innovative research results of this paper include:(1)In view of the weak interactivity of traditional time-varying structural mode mining methods,which affects the time complexity and robustness of mining time-varying structural modes,a hyperbolic time-varying structural mode visual mining method based on polar coordinate space is proposed.Principal Component Analysis(PCA)method is used to map the energy features of GPR data to polar coordinate space,and human-computer interaction technology is used to analyze,explore and mine hyperbolic time-varying structural modes from the data.The method utilizes the user's domain knowledge,visualization technology and interaction technology,allowing users to quickly mine hyperbolic time-varying structural modes from complex GPR data.The case study results show that the method effectively improves the robustness of the mining process and reduces the time complexity.(2)Aiming at the problem that the traditional time-varying structural modal mining method has a weak ability to learn data features,which affects the robustness of mining time-varying structural modalities,a hyperbolic time-varying structural modal detection method based on visualization optimization is proposed.The hyperbolic wave time-varying structure mode detection model VGG-GPR is constructed by transfer learning method and VGGNet network model,and then the model-insensitive data is mined by Largevis-based visual data mining method,which is used as a new training set to train the model again to optimize the performance of the model.With the advantage of visual data mining,this method enables users to mine data that is not sensitive to the model quickly and accurately,thereby improving the performance of the model.The experimental results show that this method can effectively improve the robustness of the hyperbolic time-varying structures modal detection process.