Authors: Fan Shi*, University of Texas At Dallas, Fang Qiu, University of Texas at Dallas
Topics: Remote Sensing, Transportation Geography
Keywords: Traffic monitoring, remote sensing, video
Session Type: Paper
Start / End Time: 3:20 PM / 4:35 PM
Room: Plaza Ballroom F, Sheraton, Concourse Level
Presentation File: No File Uploaded
The availability of spaceborne video sensors has offered remotely sensed videos to observe the Earth surface, and traffic monitoring is one of the potential applications. However, there is very little research on this subject, and current achievements are challenged by two problems. First, after moving objects are detected, it is difficult to remove false alarms located outside the road network. Second, although moving vehicles can be detected, they cannot be successfully tracked with the spatial resolution of video data. This study was carried out to solve both problems by modeling the road network and traffic flow using the remotely sensed video. Specifically, the road network was modeled by accumulating trajectories of moving objects so that any detected moving objects outside the road network were treated as false alarms. Then, the traffic flow was modeled by tracking moving vehicles based on the spatial proximity theory. As a result, dynamic properties of moving vehicles like the velocity, longitudinal and lateral position, etc., were extracted. The proposed methods were tested on a greyscale video acquired by a Skybox satellite and the results perform very high accuracy compared to the ground truth.