Intelligent traffic monitoring systems need automated mechanisms to detect incidents at urban intersections. In this paper, a computer vision based incident detection system has been proposed which works on feature detection and tracking over consequent frames. Among the several feature detectors, harris corner detector has been applied on the incoming image stream and tracked using correlation approach. These feature points are used to obtain the velocity information of the vehicles, which is fed to a change point monitoring (CPM) algorithm. The CPM scheme monitors the changing velocity from one frame to the other and triggers a change whenever a significant change in data distribution is observed. This paper uses the ADaptiveWINdowing (ADWIN) technique, which is able to detect the change in sequential data at every time instant. The proposed framework has been applied on the TRIMARC dataset, which has been captured at the intersection and is found to work promisingly. © 2017 IEEE.