Research On Underground Pipeline Simulation And Detection Based On Ground Penetrating Radar

With the advancement of infrastructure construction,people pay more and more attention to the harm caused by the removal of old pipelines.For the detection of targets in underground pipelines,ground penetrating radar is a very efficient and accurate method.However,ground penetrating radar generates a large amount of data,so there is an urgent need for efficient and accurate automatic target detection technology.Nowadays,the related fields of artificial intelligence are paid more and more attention,and the neural network has great advantages in image target detection,and is also a hot topic in image target recognition and detection.According to the above problems,this article will conduct in-depth discussion and analysis from three aspects,namely the design and implementation of hardware equipment,namely the design and implementation of high-speed ground penetrating radar data acquisition board,underground pipeline simulation and convolution neural network based underground Pipeline inspection algorithm.The main research content and innovative work of the paper are as follows:1.Design and implement a multi-channel ground penetrating radar high-speed data acquisition board.The ground penetrating radar acquisition board implemented in this paper runs in a multi-channel parallel working mode through a channel switching module,and finally obtains a three-dimensional ground penetrating radar image.At the same time,the purpose of equivalent sampling can be achieved through the cooperation of the acquisition module and the delay module.2.For the simulation of underground pipelines,in this paper,a large number of2 D and 3D underground pipeline models are generated in advance by writing programs,and then FDTD calculation is performed on them by Gpr Max software,and finally a ground penetrating radar echo file is obtained.After imaging the echo file and a variety of data augmentation image processing methods,a large number of GPR sample data are obtained.3.For the three-dimensional ground penetrating radar image,this paper proposes to take screenshots in multiple directions to convert the three-dimensional target detection problem to two-dimensional target detection,and select the appropriate cross section in each direction.This paper uses image information entropy to select.In this paper,based on the convolutional neural network,the target detection of the underground pipeline of the two-dimensional ground penetrating radar image is realized.On this basis,a parallel convolution network is constructed,and the cross section in the three directions of the ground penetrating radar image is used as input to realize the three-dimensional ground penetrating.Radar image target detection for underground pipelines.