Data fusion for accurate characterisation of buried cylindrical objects using GPR

The detection and characterisation of buried cylindrical objects using GPR is a highly subjective process, due in part to the highly variable nature of the data and the high degree of uncertainty associated with the target geometry and subterraneous environment. Broad assumptions are often made, significantly limiting the accuracy and hence the value of any subsequent interpretations. This problem is exacerbated when data is of poor quality or when site coverage is sparse. By combining information extracted from various cues from within the data in a manner that minimises the reliance on ready-made assumptions, rules of thumb and conjecture, it is possible to appreciably enhance the reliability and accuracy of the final interpretation result. Factors such as the permittivity of the medium, the radius of the cylinder, both the azimuth and inclination angles of the cylinder, as well as the electromagnetic radiation pattern of the antennas, are taken into account. This data fusion is performed in an automatic and unsupervised fashion across an entire site, in conjunction with neural network-based segmentation procedures. The result is a three-dimensional interpretation report that can be readily visualised without the need for specialist knowledge.