Geotagged Traffic Video Analysis with WebVMT

By Rob Smith (Away Team) and OGC

Demonstration using roadside video footage with GeoPose to analyse traffic using WebVMT in a web browser, including vehicle tracking, data synchronization with video, multi-device data aggregation, wrong-way traffic identification and forensic speed calculation.

Skip ⬇

Video

Transcript

Web Video Map Tracks or WebVMT is an open format designed to synchronize location with video for the web.

In testbed 19 Away Team recorded geotagged video in a collaborative effort at Ordinance Surveys headquarters near Southampton.

Footage was recorded concurrently by two smartphones mounted on tripods at the roadside and by cameras on Ordinance Survey's Street Drone autonomous vehicle.

GeoPose for each roadside camera was recorded using the 3D compass app developed in tested 18.

The GeoPose recorded by the camera is shown here in orange and its corrected GeoPose in blue.

Detection zones were identified where camera sight lines intersected road lanes.

Video frames showing a vehicle directly in a sight line were then geotagged with a corresponding lane location using WebVMT to create trajectories synchronized with the footage.

A WebVMT utility was developed to calculate the closest approach time of a trajectory to a given location with millisecond accuracy.

This enabled geotagged videos to be synchronized by matching the vehicle's location recorded on board with that observed by the roadside cameras.

Data were aggregated using tested 17's WebVMT tools with metrics showing that synchronization accuracy is within 100 milliseconds.

A robust method was devised to determine traffic flow direction by monitoring the order in which a vehicle passes through detection zones.

This technique can discriminate between adjoining lanes and detect unusual events such as u-turns.

Sight line analysis was also applied to CCTV footage shown in the BBC Crash Detectives program series 4 episode 1 to estimate vehicle speed from around 1 kilometer away.

WebVMT simplifies geotagged video analysis.

Sensor data has been successfully synchronized with video and interpolated.

Video footage from multiple cameras has been accurately synchronized and data aggregated from multiple sensors.

Full details can be found in the tested 19 engineering report.