Radius estimation of buried cylindrical objects using GPR — A case study

Abstract

Buried object detection is an important and difficult problem. It has significance in civilian situations like excavation of trenches without damage to pipelines to military manoeuvres in mine-laden areas. Often information on buried pipes or mines are unavailable after passage of some time and Ground Penetrating Radar (GPR) is fast emerging as an effective technique for these applications. The results, however, are not always encouraging and success depends on the combination of ground conditions, object geometry and material, antenna characteristics and, above all, on understanding of GPR response dependence on object and host medium parameters. This is an area where there is a need for further research. This study deals with detection of buried pipes. It is primarily intended to develop a satisfactory method to locate the pipes and to evolve guidelines for choice of appropriate antenna for a given location. Eight sites were carefully selected for evaluating GPR performance. These sites were selected because the buried pipes could be accessed for verification of their characteristics. These sites encompassed a range of cover materials above the pipes, pipe material types and dimensions. The cover types were paver blocks, stone slabs, concrete and bitumen. An important characteristic which helps in interpretation is the appearance of the hyperbolic frown. The approach for the pipe studies involves data collection using 200MHz and 400MHz frequency GPR antennae, extraction of the hyperbolic frown through Support Vector Machine (SVM) based classification and skeletonization. Relationships between depth to diameter ratio and hyperbola formation has been examined. Computation of the radius and depth of pipes was done by fitting a theoretical hyperbola over the observed one. Computed radius and depth are compared with ground truth values. The study shows that GPR response depends on frequency, cover material type and pipe material. As far as the hyperbola is concerned, it was seen to be good in case of paver block-concrete pipe and paver block-steel pipe combinations. Under concrete cover the results varied in quality. Under bitumen cover the results were unsatisfactory. In general, 400 MHz performed better. The study points to the need for further investigation on different pipe materials under different soil cover conditions.