Research On Hyperbolic Wave Extraction Method Of GPR Image Based On Graph Cut

Ground Penetration Radar(GPR)is a widely used non-destructive detection technology,which mainly uses high-frequency radar waves to detect underground targets and their structures.Different underground targets show different characteristics in radar images due to their different properties.This characteristic is mainly represented in the form of the hyperbola in GPR B-scan images,which is called a hyperbolic wave.Information such as the location and burial depth of underground targets can be explained by hyperbolic waves,so the extraction of hyperbolic waves plays an important role in understanding underground targets.But the interpretation of GPR data is usually done manually by humans,which is a time-consuming task and the results are highly dependent on the experience and prior information of relevant experts,which limits its application.In view of the above problems,this paper combines the graph cut algorithm to study the hyperbolic wave extraction problem of GPR images.The specific research contents are as follows:(1)A hyperbolic wave extraction method for GPR B-scan images based on Normalized Cuts is proposed.The method divides the hyperbolic wave extraction into image segmentation stage and the hyperbolic wave fitting stage.In the image segmentation stage,mean filtering and bilateral filtering are used to remove most of the noise of the GPR B-scan image,SEEDS algorithm is used to generate superpixels with well-fitted edges,and superpixels are used as vertices of the Normalized Cuts algorithm to reduce the running time of the algorithm and improve the edge fitting rate of image segmentation.The similarity matrix is constructed by the gray value and gray distribution difference between superpixels to improve the accuracy of segmentation.Finally,the GPR B-scan image is segmented based on the Normalized Cuts algorithm.In the hyperbolic wave fitting stage,the hyperbolic wave is fitted by the RANSAC algorithm.The experimental results show that this method can not only improve the accuracy of hyperbolic wave segmentation,but also accelerate the speed of hyperbolic wave extraction.(2)A hyperbolic wave extraction method for GPR B-scan images based on Grab Cut is proposed.The method divides the hyperbolic wave extraction into image segmentation stage and the hyperbolic wave fitting stage.In the image segmentation stage,the mean filter and bilateral filter are used to remove most of the noise of the GPR B-scan image,the GB VS algorithm is used to locate the hyperbolic wave target region,and the hyperbolic wave target region is used as the initial target of the Grab Cut algorithm,which makes it unnecessary to manual annotation can automatically identify the hyperbolic wave target region.The segmentation accuracy is improved by removing the non-hyperbolic regions in the Grab Cut algorithm by using the KSW entropy algorithm and the Laplacian operator.Finally,the GPR B-scan image is segmented based on the Grab Cut algorithm.In the hyperbolic wave fitting stage,the hyperbolic wave is fitted by the RANSAC algorithm.The experimental results show that this method can not only speed up the extraction of hyperbolic waves,but also has better robustness in extracting hyperbolic waves.