Intelligent Extraction And Visualization Of GPR Hyperbolic Waves Based On Time Series Analysis

Ground-penetrating radar is a technology that uses high-frequency electromagnetic waves for subsurface detection,and is widely used for its advantages of non-destructiveness,high resolution,fast imaging,and intuitive results.Hyperbolic wave is a key morphological feature for subsurface target identification as well as location,structure,size and other important parameters acquisition,and is a form of storing ground-penetrating radar data base information.However,in practical application,due to the influence of non-uniformity of geological structure,electromagnetic interference,electromagnetic wave propagation attenuation and other factors,the hyperbolic wave presents diverse forms such as blur,confusion and discontinuity.The existing research methods often only address the specialized morphology of hyperbolic waves,and when applied to the real environment,the generalization ability of the model is weak and the time cost is high,which leads to the poor robustness of the algorithm.The presence of a large number of interfering factors in radar data poses three major challenges for hyperbolic wave extraction: first,how to reduce the dependence of the algorithm on image quality in the preprocessing process;second,how to quickly and accurately acquire the hyperbolic regions contained in the image from a large amount of data;and third,how to design a highly applicable and wide-ranging extraction method for hyperbolic waves with irregular morphology.In this paper,based on the time-series analysis,the correlation between adjacent data channels in ground-penetrating radar data is deeply explored,and the hyperbolic wave regions are quickly extracted using the vertex localization method.For the hyperbolic waves with diverse morphologies,an adapted waveform extraction model is constructed for unified modeling analysis.The specific research contents and innovation points of intelligent extraction and visualization of GPR hyperbolic waves based on temporal analysis are as follows.(1)A hyperbolic wave peak extraction method based on time series analysis is proposed.The radar data are analyzed from the perspective of time series,and the dynamic window is constructed by using the time series features,which is used to extract the key information of each data channel.The peak points extracted by the dynamic window are useful for analyzing the shape characteristics of hyperbolic waves,and laying the foundation for region of interest extraction and similarity calculation between adjacent data channels.The experiments compare five other preprocessing methods,and the method has more complete and accurate preprocessing results for fuzzy,vestigial and irregular hyperbolic waves.(2)A prototype network-based hyperbolic wave ROI intelligent extraction method is proposed.Based on the extraction of peaks based on the time sequence,a vertex localization method based on shape features is designed for acquiring regions of interest and making full use of the distance between hyperbolic wave samples and non-target samples for target classification.The vertex localization method can quickly locate the local area containing hyperbolic waves and avoid searching the whole area of the profile,which reduces the time complexity of the algorithm.In order to adapt to the situation of ground-penetrating radar images with cluttered background and more clutter,the attention module is added to the hyperbolic wave classification process so that the key information in the image can be better explored,and thus the cluttered regions can be filtered quickly.Experiments show that the method has good stability and accuracy for ground-penetrating radar region of interest extraction.(3)A hyperbolic wave extraction and visualization method based on peak-point similarity fitting are proposed.The method analyzes from a timing perspective,especially for the problem of hyperbolic waveform morphological breakage in images,constructs a waveform extraction model,and uses the similarity of subwave regions to obtain the set of peaks of hyperbolic waves.The clutter waves are effectively separated from the target hyperbolic waves by fitting,and the algorithm's dependence on the image quality is reduced.The extraction process follows a search-and-extract approach,which further improves the efficiency of target detection and facilitates visual modeling analysis later.During the experiments,the method can improve the robustness of the algorithm under complex background and clutter interference environments.