A Knowledge-based System for Pedestrian’s Roadway Crossing Behavior through Video Cameras

An attempt was made to investigate behavioral responses for pedestrian crossing roadways using a normal-based camera setup and a Personal Computer (PC)-based vision system along with an expert system developed specifically to help non-experienced people to perform a safe roadway crossing. This process was demonstrated by studying conflicts between pedestrians and vehicles as an indicator for a pedestrian crash. Two normal-based cameras were used to film pedestrian-traffic movements. A vision system was used to extract about 3317 conflict observations through digital images at six different locations in Irbid-City, Jordan. A database of pedestrian, traffic and geometric related information was developed. The collected variables in this database included: pedestrian’s speed, vehicle’s speed, vehicle distance, type of vehicle, geometry of the road, land use, location of conflict, pedestrian facility, pedestrian distance from the crossing location, age of pedestrian, gender of pedestrian and angle of crossing. Statistical regressions were carried out to establish useful models to estimate pedestrian’s speed from the mentioned variables. An expert system with the basic If... Then forward and backward chaining of the knowledge-based rules along with decision trees was developed using the VP-Expert Shell in order to help nonexperienced pedestrians in making safe decisions to perform roadway crossing. The system was validated and checked with actual data of pedestrians crossing in different locations for both: safe crossing and not crossing cases. Results of this investigation indicated that: 1) The linear multiple regression model was the most significant model to predict the relationship between pedestrian’s speed and the developed database variables; 2) Vehicle’s speed, gender of pedestrian, distance between vehicles, geometry of the road, land use and location of the road, and pedestrian’s facility variables were found to be the most significant contributors to pedestrian behavior while crossing the road; 3) Normal-base camera setup has proven to be a useful, practical and accurate camera configuration and data acquisition system for pedestrian and traffic studies; 4) Conflicts between vehicles and pedestrians can be used as an indicator for pedestrian crashes; and 5) Expert systems have proven to be useful educational systems to assist non-experienced pedestrians to perform decisions regarding safe roadway crossing.