Traffic monitoring at signal-controlled intersections and data mining for safety applications

Signal-controlled intersections in urban settings represent significant roadway junctions where traffic flows accumulate and intersect, and where accidents tend to concentrate due to the crossing paths of vehicles potentially in conflict. For intersection safety solutions, it is essential to gather the traffic flow information in real time so that proper advisory warning can be issued or control actions can be taken to minimize hazards and to avoid collisions. This work presents an exemplar set of data mining techniques based on real-world traffic monitoring. The methodology is illustrated by two case studies, collision warning for left-turn across-path scenarios and signal violation. The use of these techniques can yield valuable information applicable to a variety of roadway safety and traffic management applications.