3D-GPR-RM: A Method for Underground Pipeline Recognition Using 3-Dimensional GPR Images

Ground penetrating radar (GPR), as a non-destructive and rapid detection instrument, has been widely used for underground pipeline detection. However, the interpretation of 3-dimensional GPR images is still manually performed, process inefficient. Aiming at solving challenges automatic recognition pipelines, we propose method based on deep learning algorithm, which uses improved 3D depth-wise separable convolution block. In order to expand number samples in dataset, data augmentation three-dimensional matrix rotation use wavelet-based denoising filter out direct wave interference. To prove effectiveness efficiency our method, compared classification performance convolutional block with traditional ordinary under same conditions. According experiment’s results, parameters proposed 66.9% less than that while similar. Furthermore, convolution, can significantly improve ability neural network, calculations remain almost same. This study demonstrates 3D-CNN field image interpretation. An also proposed. It greatly reduces amount calculation ensuring performance. better existing algorithms At time, obtain position direction pipeline, this study, conic fitting using Canny operator locate vertices B-Scan record their horizontal vertical coordinates. estimate it lays foundation future work such measuring depth.