Estimation of traffic flow using sparse camera images from LTA

Abstract

With the increased number of vehicles on the road in Singapore, traffic congestion on expressways is getting more common. Motorists have to rely on EMAS, a system that provides traffic alerts by LTA. However, the system presents the traffic alerts on the road itself, and it may be too late for motorists to switch to alternate routes to their destination to avoid the congestion.

By employing an image processing algorithm to assign road condition type to images taken by the cameras on the expressways and then displaying the results on the web browser for motorists to view before taking their trip, motorists can now plan their driving route to avoid roads that are heavily congested.

The Model-View-Controller architecture is used. Two programs were created; the first program deals with extracting the webcam images from LTA’s database and storing them into a MongoDB collection, while the second program processes the webcam images and then assign a road condition type to the images. The processed images along with its assigned road condition types are then displayed on the web browser with Google Maps.

Three algorithms namely, (1) Manual, (2) Percentile-based and (3) K-nearest neighbours, were used and experimented to find out which algorithm provides the highest accuracy in estimating and assigning the correct road condition type to a test image. The Percentile-based algorithm has the highest accuracy of 80% in assigning the correct road condition type.

Given that the algorithms can be affected by poor image quality due to external factors, some improvements on the algorithms have been proposed and could be used in the future. A mobile application can also be implemented given that motorists may request such traffic information on the go.