Automatic detection model for underground pipelines using FDTD analysis and convolution neural network

초록

As the demand for underground space continues to grow, the importance of obtaining information on underground facilities has become crucial. With recent advancements in deep learning techniques, there has been an increasing interest in employing deep learning models to discriminate GPR signals. However, the performance of these learning models heavily relies on the quantity and quality of the training data. Utilizing numerical analysis techniques can serve as a viable alternative as it allows the accumulation of data encompassing various materials and conditions. Thus, this study focuses on training using simulation data from heterogeneous ground and compares the performance of the learning models. In conclusion, this study presents a novel approach to discriminate GPR signals using numerical analysis and deep learning techniques. By training the models with simulation data from heterogeneous ground, a more comprehensive understanding of the performance of the learning models can be achieved.