Underground Voids Redognition Method Of Ground Penetrating Radar Based On Low Rank And Sparse Theory

With the development of urbanization in China,road surface collapse accidents caused by underground engineering construction occur frequently.Therefore,it has become an important task in urban construction to detect urban roads in time,reduce disasters and ensure road safety.Ground penetrating radar(GPR)is a technique which uses electromagnetic wave reflection to determine the distribution of underground medium.It is widely used in road disease detection because of its advantages of nondestructive,fast and high resolution.Underground cavity is the main cause of road collapse.Therefore,it is of great practical value to study underground cavity detection based on GPR.The radar echo data is seriously polluted due to the complex composition of underground medium,external electromagnetic wave interference and strong clutter,so it is necessary to analyze and interpret the echo data.At present,the interpretation of GPR data depends on expert experience,so the speed of data analysis is slow and the misjudgment rate is large.Therefore,it is urgent to study the automatic identification method of GPR underground cavity data and to realize the automatic processing and interpretation of radar echo data.In this paper,a detection and recognition method based on low rank and sparse theory of GPR underground cavity is proposed.This method combines robust principal component analysis and sparse representation classifier to realize the automatic recognition of underground targets.The validity of the method is verified quantitatively by simulation and experimental research.The main research work includes:1.Summarize the existing methods of underground cavity detection,introduces and analyzes the application of ground penetrating radar in underground cavity detection and the research status of detection and recognition methods of GPR targets.2.Introduce the working principle,system structure,performance parameters and data form of GPR.3.A method of penetrating radar clutter suppression based on low rank sparse theory is proposed,and the corresponding experimental system is built.The simulation and experimental research are carried out for human body target detection and underground cavity target detection,respectively.The validity of robust principal component analysis(RPCA)method for clutter suppression is quantitatively verified.The experimental results show that compared with the traditional mean filtering method and principal component analysis method,this method can effectively improve the signal to clutter ratio and has a better clutter suppression effect.4.In this paper,an underground target recognition method based on low rank sparse theory is proposed.The robust principal component analysis is used to suppress clutter,then FastPCA is used to extract features.Finally,sparse representation classifier is used for target recognition.The method is applied to mine target experiment data,simulation and experimental data of underground cavity,respectively.The results show that the proposed method can automatically recognize underground targets and has a higher recognition rate than the traditional support vector machine classification method.