GNGTS 2019 - Atti del 38° Convegno Nazionale

700 GNGTS 2019 S essione 3.2 and Annan, 1989). At the end of ‘90’s the “attributes”, defined as a specific measurements of geometric, kinematic and dynamic or statistical features (Chen and Sydney, 1997), firstly used for reflection seismic data analysis, were exploited also toGPR data analysis (Young et al. , 1997). The main objective was improve the affordability of interpretation and detect objects, structures and parameters that are not clearly visible in the original data, while limiting, erroneous results. The best achievements refer to archeology (Zhao et al. , 2013; 2016a), glaciology (Zhao et al. , 2016b), structural geology (McClymont et al. , 2008; Forte et al. , 2012), and stratigraphy (Zhao et al. , 2018). Less relevant applications are reported for engineering, particularly as far as underground utilities location. This work aims to verify if GPR attributes can be exploited when applied to actual 3D GPR volumes for a faster, more reliable, and semi quantitative characterization of buried pipelines in terms of both location and geometry. Specifically, we evaluated the applicability of 14 attributes in order to determine which of them would provide the best results and would be most suitable to map in semi-automatic mode underground utilities. For semi-automatic interpretation we adopted a strategy similar to the one described in Dossi et al. (2015), but based on a few picks provided by the interpreted (seeds) then automatically extended to the whole GPR volume. Methods. We tested the proposed methodology on a 3D GPR dataset acquired in 2017 by Anfibia srl in a parking near Ravenna, with a MiniMIRA8 3D GPR (MALÅ-Guideline Geo) system. This equipment has shielded antennas with 400 MHz central frequency, operating as 8 antenna pairs and so collecting 8 parallel, 8 cm spaced GPR profiles. We set the trace interval to 8 cm, equal to the profile spacing, thus collecting a “ full-resolution ” (Grasmueck et al. , 2005) dataset, which overcomes errors and limitations of large-spaced bidimensional profiles (2.5 D). Moreover, 3D data acquisition makes almost irrelevant the importance of direction of the acquisition respect to the orientation of buried targets, which instead is challenging in usual bidimentional surveys. As previously highlighted, the main objective of this work was to verify the applicability of GPR attributes calculated on 3D dataset in order to semi-automatically interpret buried pipelines. In this regard, both acquisition and processing were carried out in standard mode, because they don’t represent the primary focus of the research. Therefore, the collected data volume was processed by applying background removal, band pass filtering and amplitude correction by means of spherical divergence and exponential corrections. A Stolt 3D time migration was further applied adopting a smoothed velocity field derived by exemplary diffraction hyperbolas analysis. At this point, different attributes were calculated on the entire data volume by adapting for GPR Petrel interpretation suite (Schlumberger). Attributes are related with different quantities measured on seismic or GPR data including amplitude, phase frequency, texture, and many others. In this study we focus on volume attributes calculated on the entire 3D GPR dataset. Specifically, we tested 14 attributes, pertaining to 4 different categories: - Signal attributes (first derivative, second derivative) - Complex attributes (apparent polarity, cosine of phase, envelope, instantaneous phase, instantaneous frequency, dominant frequency, sweetness) - Geometrical attributes (dip deviation, local dip, local smoothing, variance) - Texture attributes (chaos) For each attribute, calculation parameters were set up in a dedicated procedure in order to maximize their performance. This paper presents results achieved by the calculation of two attributes (first derivative and envelope) that allow the best performances in the test dataset and that match the objective of this work. The first is defined as the time rate change of the input trace. The second one (often referred as “reflection strength”) is define as the total instantaneous energy of the entire analytical trace. The adopted auto-tracking function consists of three fundamental phases: - to calculate different attributes on the entire volume, evaluating which are the attributes that are able to highlight the exact location, thickness and variations of depth of underground