Research On Feature Extraction And Application Of GPR Image Based On Multi Frequency Scale Analysis

Ground penetrating radar(GPR)reflects the underground scene through the change of electromagnetic characteristics caused by the change of electromagnetic wave in the medium.Compared with other traditional detection methods,the advantages of GPR are mainly reflected in several aspects: 1)it has high detection efficiency and can be applied to a wide range of underground detection.2)The detection method of GPR does not need to excavate and destroy the ground.It is a nondestructive testing method.3)It can work around the clock and has low requirements for the working environment.With the iterative update of GPR technology and the development of big data technology,the application of GPR has gradually moved from military to civil fields,including underground target detection and recognition,autopilot positioning and ground maintenance.This paper focuses on the feature extraction method of GPR image,and puts forward the concept of vector symmetry model based on multi frequency analysis.At the same time,based on the proposed feature extraction algorithm,an innovative algorithm is proposed to improve the application performance of GPR image matching and target detection.The research of this paper is as follows:The signal model of GPR is studied and discussed,which is extended from onedimensional A-scan signal model to two-dimensional B-scan model,and the electromagnetic characteristic parameters in different media are discussed.Combined with the actual collected data,the feasibility of this study is effectively verified.The common feature extraction models of GPR images are studied,and two feature extraction models based on multi frequency scale analysis,phase consistency and phase symmetry,are introduced.The GPR image features with low physical resolution are extracted by multi frequency scale analysis.The traditional algorithms for GPR image matching and target detection are studied and compared,and their shortcomings in GPR image are analyzed combined with the experimental results.The innovative work of this paper is as follows:1)In this paper,the concept of vector phase symmetry is proposed,and the scalar feature of symmetry is extended to the vector feature model to realize the flexible application of phase symmetry in computer vision tasks.2)For the matching task of GPR image,using the rich line features of GPR image,this paper proposes a point feature matching framework based on edge features,and proposes an edge feature extraction algorithm based on improved phase consistency to extract more robust edge features.The effectiveness of the proposed algorithm is verified by experiments on Simulation and measured data.3)For the target detection task of GPR image,this paper constructs a new target feature extraction algorithm based on the proposed vector symmetry feature model to improve the shape description ability of GPR image target detection.Compared with the popular target detection algorithms in recent years,the proposed algorithm shows an obvious performance improvement.