The Target Detection And Radius Estimation Of GPR Images

In recent decades, Ground-Penetrating Radar, as a new non-destructive detectionequipment, is widely used in various fields of national economy, and has shown agrowing interest in research area. The extraction and auto-detection of the GPR targetobjects has been one of the difficult and critical issues in the practical application ofGPR.In this paper, we first introduced the basic theory of GPR and the characteristicsfor the three data formats: the A-Scan, B-Scan and C-Scan. Then we combined withGprxMax2D based on FDTD and Matlab software to acquire the GPR modeling data,which is as the GPR original data that we utilized to verify our method for thedetection and localization of GPR target.1. In the paper, a method based on curve fitting is proposed by using thehyperbolic echo characteristics. Firstly, the GPR raw data has beenpre-processed by image processing methods: remove the surface clutter basedon the average method, eliminate the background of the GPR image by theimage segmentation, which greatly reducing the follow-up processingcomputation. And the detection for the hyperbolic edge of GPR images isdone based on Sobel operator, obtaining the thinning edge of the hyperbola inGPR images by the gradient magnitude. Then hyperbolic feature has beenenhanced for curve fitting. Finally, by using the hyperbolic symmetry, thehyperbolic echo apex positions in the GPR images are extracted and markedon their apexes. The experimental results show that all targets are detectedcorrectly with a smaller amount of calculation.2. The estimation algorithm for the radius of pipe objects is proposed in thispaper. Based on the hyperbolic features, the parameter information of thehyperbola is firstly obtained by the curve fitting based on direct least squaresmethod, and the radius of the target is estimated with small errors. Then in ourpaper, we propose a novel method to estimate the radius of the pipe objects.Suppose that there is a GPR point-object, and the information has been obtained. By using the prior information, we can compute the radius of thetarget that is buried in the same depth and in the same reinforced material, andthe errors is smaller. Furthermore, the method has been extended to estimatethe radius of the unknown object based on prior known targets information.The results also show that the method can not only accurately calculate thesize of the target radius, but also has the advantage of simple operation andthe computation is relatively small.