Landmark selection preferences of young students under orientation task within street environment

ABSTRACT Landmarks are important spatial reference in orientation because they can help pedestrians understand the relative spatial relationship between environment and themselves, yet not all landmarks are suitable for orientation. Pedestrians usually select the proper ones based on their own cognition, which is a time-consuming process. If we can predict popular landmarks (preferred by most people) from a first-person perspective, orientation will be much easier. To better understand landmark selection behaviour, the research in this paper designs an orientation experiment within virtual street environment to investigate young students aged 21–31 about preferences for landmarks. Based on the selection results, this paper further constructs a random forest model to choose popular landmarks. Results indicate that one or two landmarks are enough to meet young students’ needs for orientation, and the time spent on orientation significantly reduces by using chosen popular landmarks in most scenes. In addition, classification model achieves a desirable performance in popular landmark selection: F1 score and AUC of the predictive results reached 0.820 and 0.855, respectively, where landmark route deviation, visual salience and semantic salience serve as essential factors in influencing young students’ selection. These findings can provide a feasible reference for optimising pedestrian navigation system.