The vibrations caused by moving wheeled vehicles can reveal much about vehicle location, velocity and size.
As a data-source they are furthermore anonymous and weather-proof. So it is perhaps surprising that little empirical knowledge exists about the seismic footprint of traffic.
We had a serendipitous opportunity to look into this "terra incognita" in 2014: an oil company decided to share with us their seismic imaging data consisting of 5200+ sensors spaced about 90m apart and blanketing Long Beach (CA). Such exceptional data may actually become less exceptional in the future: changes in sensor costs, weight, and power-usage are already causing a trend towards more dense and distributed sensor networks in seismic studies.
I’ll start this talk with a quick excursion on the seismic remote-sensing paradigm. I’ll then describe some existing work on seismic effects from vehicles and the use of seismic sensors as in-roadway traffic counters. Of course we’ll also peak into the treasure trove of the Long Beach data to see how various analytical tools such as spatio-temporal filtering, mixture models, network theory, and sparse matched-field processing can be used to reveal traffic information.
Back to Long Programs