Classification of buried objects from series of aligned hyperbolic arcs or "pendants" in radar scans: Estimating a buried pipe diameter

In our experiments, 600 MHz radar scans across long parallel objects, such as pipes, buried in one metre or so of soil, show complex reflection patterns consisting of a series inverted hyperbolic arcs characteristic of the object. A classification of the objects has been achieved by an analysis of the arcs, which gives rise to a series of "apex" points defining the lateral position (y) and depth (z) and amplitude of each arc. For objects of size larger or comparable with the wavelength (20cm), several points with alternating positive and negative phases are obtained. Code has been written to associate series of apexes which may all arise from the same object. For example these should all lie within a specified vertical area, and which have appropriately spaced depths, with each ripple having the correct alternating phase. The relative intensities of these apexes provide the necessary features for classification by, for example, a neural net. The method is simulated using two-dimensional examples provided by pipes buried under a road. Different pipes can be identified and readily separated from small objects giving background-scattered signals.