A distinctive similarity rendering approach to reconstitute hyperbola apices in GPR images

In ground penetration radar (GPR) B-scan images the detection of hyperbola signatures from the target object reflections is necessary, so that buried target object can be identified accurately. In order to accurately recognize the hyperbola signatures, a novel approach called a Distinctive Similarity Rendering Approach (DSRA) is applied to GPR images. This approach includes five steps. First, the GPR B-scan image is pre-processed by pattern profile method to extract the symmetrical open curves for cone presence. Second, the detected open cone curve components were enhanced by target focussing technique to build the close cone curve signatures to result the best hyperbola signature fit. Third, the processed GPR B-scan image results the best fit hyperbola signatures, these signatures are further processed using our proposed similarity rendering approach, to reconstitute the further hyperbola apices and for the detection of all hyperbola signatures in GPR B-scan image. Fourth, then we use sample hyperbola shape signatures with varying hyperbola cone radius (Cr) to locate the hyperbola apices from the spurious reflections of all hyperbola region of interest detected. To find the hyperbola apices, a Hough transform is used in the similarity rendering area. Finally, we estimate all the hyperbola shape signatures of target object reflections. In terms of hyperbola apices reconstitution, the proposed approach exhibits better performance, as shown in experimental result explanation.