Multi-dimensional Detection of Underground Utilities Using a Data Mining Method

Accident breakage of unknown utilities (e.g., cables and pipes) frequently occurs during underground construction and causes temporary cut-off of energy and water supplies in urban cities. Mitigation of the excavation risks requires fast and accurate detection of subsurface utilities, but the current geophysical methods suffer from high computational cost and human subjective judgement. Here we developed a data mining method for fast detection and location of underground utilities from the collected data using the seismic scattering theory. The detection of underground utilities is based on analyzing the differences in relative energy levels of effective signals among various seismic records that are generated by a moving seismic source. Specifically, when the position of a seismic source is close to underground utilities, the energy of effective signals increases. The increased intensity is proportional to the size of the utility. The energy change is captured using data mining algorithms including scale analysis and low-rank feature extraction. Finally, underground utilities can be identified by a larger anomaly score which is defined by the energy of extracted effective signals. The effectiveness of the proposed method has been demonstrated by testing on field underground water pipes detection.