Using Seismic Data to Build a Deep Learning Model for Detection and Characterization of Vehicles

In industrial settings, vehicles play a crucial role in facilitating operations involving the movement of materials, equipment, and personnel. Traditionally, video surveillance has been employed to monitor and track vehicle traffic at facilities. However, concerns over privacy and susceptibility to light or weather conditions have prompted the exploration of alternative technologies. Seismic sensors, due to their unobtrusive nature and ability to capture ground motions generated by vehicle movements, offer a promising solution to complement video surveillance. We assess the feasibility of using seismic data to augment existing technologies for vehicle detection and characterization. A data-driven workflow was developed to enhance the generalizability of using seismic data for vehicle detection. By employing unsupervised machine learning algorithms, we analyzed continuous seismic data collected at an industrial facility in Texas to automatically categorize the data based on signal similarity. One of these signal groups exhibited a temporal distribution that aligns with the work schedule at the site. To validate the workflow, we deployed three nodal-type three-component seismic stations and collocated them with an existing camera installed near a road at the main campus of the US Department of Energy’s Oak Ridge National Laboratory (Oak Ridge, Tennessee). The three geophones were deployed in a linear configuration parallel to the road and positioned within a few meters of the camera. The camera has automatic image-based vehicle detection and characterization algorithms, which provide a valuable dataset to validate our results. We used unsupervised machine learning algorithms to automatically generate labels using continuous seismic data. Subsequently, we trained a deep learning model designed to categorize whether a seismic signal contains a vehicle-related signature. Our model achieved an impressive F1 score exceeding 0.9. These findings underscore the potential of seismic sensors as a valuable complementary tool for vehicle monitoring, offering a resilient and privacy-conscious alternative to camera-based systems.