Investigation of Underground Anomaly by Application of Convolutional Neural Network for Ground Penetrating Radar Data Analysis

Abstract

Ground penetrating radar method is widely known for its effectiveness in investigation of underground civil engineering structures. They could be represented by the high-frequency electromagnetic signals as reflection or diffraction events in the measurement data slices. Meaningful high-frequency electromagnetic wave signals are reflected/scattered from the underground objects. Requirement of fast locating the underground anomalies was an inspiration to apply modern technology of artificial intelligence in the ground penetrating radar data analysis. There is suggested a novel workflow for detecting diffracted signals which uses the convolution neural network for this research paper. The real high-frequency datasets measured in the Ho Chi Minh City area, Vietnam are used as training, testing, and validating datasets for building a convolutional neural network model. The measured data in Nguyen Van Cu Street area, District 5, Ho Chi Minh City, Vietnam is predicted with the network model for the high accuracy result.