Accurate prediction of the spatial location of underground pipelines is crucial to ensure the safe operation of pipelines. However, the traditional ground penetrating radar (GPR) relies too much on the image data quality and the detection personnel's experience. It can only estimate the approximate location of the underground pipeline. The acquisition of electromagnetic wave velocity is an essential part of accurately interpreting the spatial location of underground pipelines. Therefore, this paper's research proposes a data processing flow that can be applied to multi-offset GPR, including multiple stacks of the same target and interactive velocity spectrum. The accuracy and practicability of the method are verified by numerical simulation and field tests. The research shows that multiple stacks can effectively suppress noise and enhance the signal characteristics of deep targets. In the simulated data, traditional GPR distance errors for layered structures and pipelines are 0.06m and 0.13m. In comparison, the distance error of the method proposed in this paper is 0.019m in layered structures, which is a reduction of 5.86% in relative error. The pipeline distance error is 0.029m, and the relative error is reduced by 10.1%. The distance errors of traditional GPR and multi-offset GPR for pipelines in field tests were 0.15m and 0.093m, respectively, with a reduction of 8.77% in the relative error.