This is the fourth history blog in a six part series – Leapfrog’s fast RBF. If you missed part three, Modelling in full 3D, you can find it here.

In 1999 work began on new surfacing algorithms so that an RBF implicit model could be utilised by conventional computer graphics packages. This meant converting an implicit RBF model into meshes of triangles and piecewise continuous spline surfaces.

The main difference between traditional RBF’s and what became ARANZ’s FastRBF™ is the ability to deal with large datasets of well over 1,000,000 points on ordinary computing hardware incredibly quickly. The maths used to speed up the calculation was initially used in particle physics. Filtering and approximation methods make Fast RBF™ ideal for visualising and processing non-uniformly sampled noisy data. FastRBF™ has extraordinary extrapolation capabilities, even when large gaps occur in a data set.

Inherent ability

Rick describes it as being like the old French curve rule, “The functions give you a very natural curve, that doesn’t bend any more than it has to fill the gaps. With the RBF you’d look and think a really good artist did that but it’s just its inherent ability.”

Continues Rick Fright, “Further refinements in FastRBF™ technology at this time were to do with making it robust to noise (errors) in the input data and being able to fit functions with different smoothness properties (E.g. Hermite and errorbar fitter).”

Interest from mining

Directors, Rick Fright and Bruce McCallum also considered new applications for FastRBF™ beyond implicit surfaces, primarily to modelling 3D volumetric data, assay or grade data are examples of such data in a mining context. They also looked at terrain or topography modelling with RBFs. These examples of FastRBFs along with some key publications in 2001 attracted the interest of a geologist from the mining consultancy SRK in Perth.

Says Rick Fright “Mining presented the perfect density modelling problem, which is one of modelling the geological strata and concentrations of minerals. We were able to do this directly in 3D. We learnt that the mining geologists were analysing the density problem in terms of creating surfaces, the boundary layers between strata, or between one concentration and another of a mineral. In other words, they were going about the job in reverse of what we had achieved, because they could not do it directly.”

Leap forward

A joint venture was formed between SRK Consulting and ARANZ called Zaparo to commercialize this technology. SRK used the early software in their consultancy with their mining customers and in these early days the geologists at SRK were the source of significant insights into what was required to move the next generation of geological modelling.

The shared vision of SRK and ARANZ to change both the methods and speed of geological modelling provided the inspiration for the name ‘Leapfrog’. With the addition of new customers outside of SRK the development of Leapfrog® now began in earnest.