Traffic State Estimation using Onboard Vehicle Camera

Traffic State Estimation (TSE) refers to the process of estimating traffic states, namely flow, density and speed in a specific road or a road network. The growing demand for TSE in cities requires real-time spatiotemporal information from a large expanse of the city. Since continuous and long-term observations in the entire network are practically impossible, the traffic states in the unobserved time-space get estimated from the partially observed data.

There have been a growing number of studies for TSE using probe vehicles, also called Floating Car Data (FCD). With the advancement of camera systems in vehicles, there is a huge potential for traffic sensing using street-level videos/images from probe vehicles for TSE at both microscopic and macroscopic levels. In the seminar we present a novel idea of using an onboard monocular camera (e.g a smartphone or action camera) to capture street-level video of the road network and gather traffic states of the observed network.