Research And Application Of Multi-sensor Underground Pipeline Detection

The underground pipeline system is an indispensable part of the operation and de-velopment of modern cities.As an important infrastructure,the underground pipeline system not only provides living materials for the residents of the city,but also provides basic resources for the production and development of the city.A well-developed un-derground pipeline system and its safe and stable operation are important guarantees and foundations for the operation of a modern city.However,due to the long service life of some underground pipelines,the existed pipeline information could be incom-plete or inaccurate.In urban constructions,due to the lack of accurate pipeline maps of the construction area and effective pipeline detection means,the latest pipeline map could be hard to get,resulting in the inability to effectively avoid pipelines during con-struction.The damage to the pipeline might affect the normal life of urban residents and the healthy operation of the city,and even causing some accidents.How to accu-rately detect underground pipelines and obtain the pipeline map of the detected area is a subject worth studying.Based on the analysis of the existed underground pipeline detection technology and related achievements,this dissertation aims to efficiently and automatically detect underground pipelines and obtain the underground pipeline map of the detected area,and carries out the research on the Ground Penetrating Radar data processing,detection of insulated pipelines in the soil,and mapping the pipelines from noisy pipeline and location data.Firstly,Ground Penetrating Radar(GPR),as a non-contact underground medium detection method,has been widely used in underground pipeline detection due to its fast speed and minimal ground intrusion.However,due to the large number of the pipelines,the amount of GPR data increases,while the GPR data interpretation is professional and complex.To address this issue,a GPR image interpreting model is proposed in this dissertation.The model could handle GPR B-scan image,extract and fit the hyperbolic features generated by the underground pipelines in the GPR image,and then calculate the depth and radius of the underground pipelines through the hyperbolic parameters.Next,since the characteristics of the insulated pipeline in the GPR image might be unrecognizable,misjudgment could happen when directly processing the GPR image to estimates insulated pipelines in the soil.Compared with GPR,the electric-field method could be more sensitive to the insulated pipeline in the soil,but the accurate information about the pipe is hard to obtain using electric-field method alone.This dissertation proposes an insulated pipeline detecting model which combines the GPR and electric-field method.The model could locate the buried insulated pipeline from the electric data,and segment the corresponding GPR image.The pipeline-independent GPR image is discarded,while the image segments that might contain the pipes are reserved and further analyzed to estimate the depth and radius of the buried pipelines.Finally,after obtaining the pipeline and positioning data of multiple detected points in the detected area,it is necessary to fuse these data to analyze and speculate the pipeline map of the detected area.The underground environment could be complex,thus the underground pipeline data obtained from the sensor could not be completely accurate,as well as the positioning data obtained by satellite positioning means or self-positioning equipment.In this paper,a pipeline mapping method based on the probabilistic mixture model is proposed,which could map underground pipelines with the largest possibility from the noisy pipeline and positioning information.The errors of pipeline and positioning information are both taken into account in the proposed method,which improves the robustness of our method in different environments.