18 Oct 2024

Automated Change Detection in Ground-Penetrating Radar using Machine Learning in R

Abstract: Ground-penetrating radar (GPR) is a useful technique for subsurface change detection but is limited by the need for a subject matter expert to process and interpret coincident profiles. Use of a machine learning model can automate this process to reduce the need for subject matter expert processing and interpretation. Several machine learning models were investigated for the purpose of comparing coincident GPR profiles. Based on our literature review, a Siamese Twin model using a twinned convolutional network was identified as the optimum choice. Two neural networks were tested for the internal twinned model, ResNet50 and MobileNetV2, with the former historically having higher accuracy and the latter historically having faster processing time. When trained and tested on experimentally obtained GPR profiles with synthetically added changes, ResNet50 had a higher accuracy. Thanks to this higher accuracy, less computational processing was needed, leading to ResNet50 needing only 107 s to make a prediction compared to MobileNetV2 needing 223 s. Results imply that twinned models with higher historical accuracies should be investigated further. It is also recommended to test Siamese Twin models further with experimentally produced changes to verify the changed detection model’s accuracy is not merely specific to synthetically produced changes.

19 Jul 2024

Permafrost and Groundwater Characterization at the Proximity of the Landfill, Fort Wainwright, Alaska

Abstract: This report summarizes a site investigation at the vicinity of the landfill, a discontinuous permafrost site, at Fort Wainwright, Alaska. The objective of this effort was to characterize the permafrost extent and groundwater flow at the study area, and to compare newly collected subsurface characteristics with historical datasets. The main tasks for this effort included lidar and remote sensing analyses, geophysical investigations, a tracer dye study, contaminant trend analysis, and installation of soil temperature sensors. Findings included changes in stream channels and watershed boundaries, and elevation losses (0.2 m to 1 m) east and northeast of the landfill. From frost probe measurements, we found that depths to permafrost were up to 1.5 m deeper in 2021 than in 2010 where the difference in depth ranged from 20% to more than 350%. Furthermore, we detected a reduction in lateral permafrost extent from geophysical datasets. The groundwater flow direction, as detected through the dye study, was south to southwest. Dye was detected up to 2,300 m from the injection point. Groundwater travel times, as calculated from the dye study, varied greatly. For upcoming historical comparisons, it is recommended that data collections are performed using similar methods as described in this study.

30 Sep 2022

Automated Ground-Penetrating-Radar Post-Processing Software in R Programming

Abstract: Ground-penetrating radar (GPR) is a nondestructive geophysical technique used to create images of the subsurface. A major limitation of GPR is that a subject matter expert (SME) needs to post-process and interpret the data, limiting the technique’s use. Post-processing is time-intensive and, for detailed processing, requires proprietary software. The goal of this study is to develop automated GPR post-processing software, compatible with Geophysical Survey Systems, Inc. (GSSI) data, in open-source R programming. This would eliminate the need for an SME to process GPR data, remove proprietary software dependencies, and render GPR more accessible. This study collected GPR profiles by using a GSSI SIR4000 control unit, a 100 MHz antenna, and a Trimble GPS. A standardized method for post-processing data was then established, which includes static data removal, time-zero correction, distance normalization, data filtering, and stacking. These steps were scripted and automated in R programming, excluding data filtering, which was used from an existing package, RGPR. The study compared profiles processed using GSSI soft-ware to profiles processed using the R script developed here to ensure comparable functionality and output. While an SME is currently still necessary for interpretations, this script eliminates the need for one to post-process GSSI GPR data.