Automated rock mass discontinuity set characterisation using amplitude and phase decomposition

Mapping structural discontinuities in subsurface environments is critical for several engineering applications. Most instances require stable rock mass to ensure safety while some require an unstable rock mass, such as in block-cave mining, for gravity induced smooth and continuous production of ore. Structural discontinuities have previously been mapped using manual, semi-automated and automated methods which have limitations in terms of required effort, processing time, definition of parameters and accuracy. Most of the semi-automated and automated approaches employ point cloud data, obtained either through structure from motion photogrammetry or laser scanner, and rely on point normal vectors in some form for discontinuity characterisation. Such methods can lead to inaccurate results in the presence of noise as normal vectors tend to show high variability in a local region. This study proposes a new automated algorithm that uses the spatial distribution of points on discontinuities to capture unique signatures in the form of a sinusoidal wave. Each discontinuity points are distinguished by unique amplitude and phase, achieved using fast Fourier transform (FFT), for characterisation in a 3-dimensional (3D) point cloud. The presented amplitude and phase decomposition (APD) approach requires minimal pre-processing and can be applied directly to a raw point cloud as filtering is inherently included in the signal processing. The method has been evaluated on an underground tunnel dataset, consisting of exposed structural discontinuity planes. The efficacy of the developed approach was tested against open-source semi-automated software (discontinuity set extractor), proprietary semi-automated software (Maptek PointStudio), manual segmentation using virtual compass plugin in open-source software (Cloud Compare), and other automated algorithmic approach based on point normal clustering and region growing. When compared against the manual discontinuity segmentation, the APD approach indicated the least error in estimating mean discontinuity dip angle and dip direction which was ±1.15° and ±1.39 with a dispersion error of ±2.24° and ±1.54°, respectively.