Recently, many researchers have been working on developing vision-based traffic monitoring systems that can collect various traffic parameters, for example, a number of passing vehicles on a road and their velocities. In this paper, a vehicle tracking method is proposed for the traffic monitoring system using a monocular grey scale camera. This method can deal with the images captured in various traffic situations such as straight road, intersection, daytime scene and night time scene, etc. To begin with, what is the best feature to detect vehicles was investigated and it was concluded that feature points are the best. The method consists of two parts: feature points detection and tracking, and vehicle tracking by clustering feature points. Experimental results reveal that the method works well for the various traffic situations. For the covering abstract see ITRD E134653.
Abstract