Integrating distributed acoustic sensing and computer vision for real-time seismic location of landslides and rockfalls along linear infrastructure

Distributed acoustic sensing (DAS) using fiber-optic cables has potential for landslide and rockfall monitoring along linear infrastructure but faces challenges for accurate seismic source localization due to signal nonuniformity and attenuation during propagation. This limits the applicability of traditional seismic location methods with DAS. We present a novel computer vision–based approach to overcome these limitations. Field experiments simulating landslide quakes and rockfall impacts were conducted near dedicated DAS arrays to validate the method. Results demonstrate the computer vision technique outperforms short-to-long-term average ratio and cross-correlation algorithms in both location accuracy and constraint of seismic sources, with locations also agreeing well with a colocated nodal seismic array. Key factors influencing performance include the type of signal processing used on the DAS data and cable array geometry. The envelope function best handled noise while L-shape and parallel dual-cable geometries proved most effective. Overall, this computer vision method provides an improved solution for seismic source location of landslides and rockfalls monitored by DAS networks, enhancing safety along vulnerable linear infrastructure like transportation corridors through mountainous terrain.