GNGTS 2019 - Atti del 38° Convegno Nazionale

GNGTS 2019 S essione 3.2 701 utilities. In this step we further considered the possibility of each attribute to attenuate (or to emphasize) the artifacts so excluding attributes that emphasized both random ad coherent noises. - to visually identify buried pipelines selecting and picking “control points” (seeds) that are surely within the targets. This step is crucial because a wrong control points selection can produce remarkable interpretation errors. - to apply the semi-automatic picking, which extends the seeds characteristics by searching for the same “signature” within the data volume. Before this operation, it is essential to define some thresholds and constraints that make the algorithm able to distinguish the target from the background. This phase, in general, is delicate and somehow subjective, but, normally, the differences between targets and background for at least some attributes is so noticeable that prevents most of false positives as well as false negatives. At the end of the above described procedure targets are picked as cubic cells with constant dimensions equal to binning of data (in the present case 8 cm by 8 cm). Possible outliers can be manually deleted in order to improve the overall quality of the results. Results. Attribute analysis . Most of the calculated attributes provided interesting information, enhancing the targets detectability in comparison with the reflection amplitude. However, some of them better visualize the underground utilities, but are not suitable for semi- automatic picking and interpretation (Table-1). First derivative attribute gives excellent results both in individuation of buried pipelines, determination of their dimensional parameters, and variations of depth. Moreover, by using such an attribute some linear artifacts related to the direction of acquisition are attenuated respect to the reflection amplitude data. Envelope also gives interesting results: targets are Table 1 - Summary table of results obtained by calculation of attributes applied to actual 3D GPR volumes. ATTRIBUTES IMAGING OF DIMENSIONAL DEPTH ATTENUATION SUITABLE FOR UNDERGROUND PARAMETERS VARIATION OF UNREALISTIC SEMI-AUTOMATIC UTILITIES FEATURES PICKING FIRST DERIVATIVE YES YES YES YES YES SECOND DERIVATIVE YES YES YES YES (partially) YES APPARENT POLARITY NO NO NO NO NO COSINE OF PHASE YES NO NO NO NO DOMINANT FREQUENCY NO NO NO NO NO ENVELOPE YES YES YES YES YES ISTANTANEOUS FREQUENCY NO NO NO NO NO ISTANTANEOUS PHASE NO NO NO NO NO SWEETNESS YES YES YES NO YES DIP DEVIATION YES NO NO YES NO LOCAL DIP YES YES (partially) YES NO YES LOCAL SMOOTHING YES YES YES YES (partially) YES VARIANCE YES YES YES YES (partially) NO CHAOS YES YES (partially) YES YES (partially) YES