Automatic incident detection along freeways using spatiotemporal Bluetooth data

Abstract

Magnetic induction loop detector technology is considered as the traditional and reliable sensors for monitoring the traffic flow at freeways. These detectors are used to monitor traffic on freeways, and many automatic incident detection models have been developed based on this detector data. However, the magnetic induction loop detectors installation is complicated and expensive. This research focuses on a special group of cross-sectional sensors, Bluetooth readers, as a “low-cost” replacement for loop detectors. In this work, a novel deep-learning algorithm was developed to automatically detect incidents in an unsupervised fashion based on Geotagged Bluetooth readings only. Preliminary experiments show promising results in correctly detecting road incidents.